Documentation · v5.3.11
ZipLoom Documentation
What every part of ZipLoom does and how to use it, checked against the product itself. ZipLoom scans AI-generated code, deploys it to hosting you own, and watches it once it is live.
Last updated: 2026-11-21 · docs v5.3.11
Deploy & Build
Getting Started
What it is. The path from zero to a first scanned, live deploy — sign up, connect GitHub, pick a repo, deploy.
Why it matters. ZipLoom sits between your repository and your host: it scans AI-generated code and gates the deploy, so nothing risky reaches production. Everything runs on a flat subscription with no usage meters (monthly, or annual at a discount).
Walkthrough
- Go to
https://app.ziploom.dev and sign up with email or GitHub.
- Set up two-factor authentication. It is required on every account (see Two-Factor Authentication).
- Complete onboarding: connect your GitHub account and grant access to the repositories you want to deploy.
- Connect at least one deploy target (Settings → Deploy Targets) — Vercel, Netlify, Cloudflare Pages, Railway, Render, Google Cloud Run, AWS, DigitalOcean, Vultr, Contabo or Amazon Lightsail — 11 in total. The accounts stay yours.
- Create a Project from a connected repo. ZipLoom auto-detects the framework.
- Trigger a Deployment. The security checks run first; a clean result deploys, an untriaged critical finding blocks it with fix guidance.
- Open the project page and read Safe To Launch? for what to fix first.
Options & Settings
- Workspace — everything (projects, deploys, team, billing) is scoped to your workspace; switch workspaces from the top-left switcher.
- Roles — Owner / Admin / Member (Settings → Team).
Edge Cases & Gotchas
- No deploy target connected → deploys can't run. Connect one first.
- GitHub access not granted to a repo → it won't appear when creating a project.
- The ZipLoom app is already installed on GitHub but this workspace shows Not connected (for example, a first attempt failed) → choose Already installed on GitHub? Link it here on the New Project page or the GitHub App card. It lists only your own account and organizations you are an owner of.
- You cannot skip two-factor authentication. Save your recovery codes when you set it up.
Related
Projects · Deploy Targets · Deploying a Project · Security (Secure section)
Last updated: 2026-11-09 · docs v5.0
Deploy & Build
Projects
What it is. A Project is a connected repository ZipLoom manages and deploys. The Projects page (/projects) lists all applications in this workspace with their framework, target platform, and last-updated time.
Why it matters. Projects are the unit everything else hangs off — deployments, scans, rollouts, and monitoring all belong to a project.
Walkthrough
- Open Projects in the sidebar.
- Use Filter projects… to find one by name.
- Each row shows: project name, Framework (auto-detected), Platform (the connected deploy target), and Updated. Badges indicate GitHub connected and the target (e.g. Vercel, Netlify, Railway) and offer Open live URL.
- Create a new project with New Project (
/projects/new) → choose a connected repo → ZipLoom detects the framework and prepares it for deployment.
- Click a project to open its detail view (
/projects/[id]) for its deployments and settings.
Options & Settings
- Framework auto-detection — set automatically from the repo; drives the build configuration.
- Platform — the deploy target bound to the project (managed under Deploy Targets).
Edge Cases & Gotchas
- A repo must be GitHub-connected and accessible to appear as a project source.
- "Open live URL" appears only after a successful deploy has produced a URL.
Related
Getting Started · Deploying a Project · Templates · Discover
Last updated: 2026-09-22 · docs v4.6
Deploy & Build
Deploy Targets
What it is. The platform accounts ZipLoom deploys into — "Connect the platform accounts ZipLoom deploys into. They stay yours." Managed at Settings → Deploy Targets (/settings/deploy-targets).
Why it matters. ZipLoom doesn't host your app; it deploys to your provider accounts. You keep ownership and can leave at any time (see Self-Host Export).
Supported Targets
Amazon Lightsail · AWS · Cloudflare Pages · Contabo · DigitalOcean · Google Cloud Run · Netlify · Railway · Render · Vercel · Vultr — 11 platforms, zero config (framework auto-detected, scanned first).
- Managed hosts: Vercel, Netlify, Cloudflare Pages, Render, Railway, Google Cloud Run.
- Servers you own (VPS): DigitalOcean, Vultr, Contabo, Amazon Lightsail. See Server Placement.
- AWS (BYOC) deploys into your own AWS account and is available on Business.
Note: Netlify here is a customer deploy target — where your app can be deployed. It is separate from ZipLoom's own hosting.
Full-Stack Apps On Cloudflare
A static site on Cloudflare deploys to Cloudflare Pages. A server-rendered Next.js app, or an Express 5 server, deploys as a Cloudflare Worker on your account, at https://<project>.<your-subdomain>.workers.dev. An Express app needs no Cloudflare files: ZipLoom wraps your server file during the build. Express 4 does not start on Workers; upgrade to Express 5 first.
- Open the project. The Next Step card at the top shows what is left: Build Workflow, then Cloudflare Setup, then Deploy.
- Set Up For Cloudflare offers two choices. Do It For Me (Owner or Admin, after you tick the confirmation) opens a pull request on your repository that adds Cloudflare's setup files. It never changes
package.json and never overwrites a file. Merge it on GitHub. I'll Do It Myself shows the steps: run npx @opennextjs/cloudflare migrate in the app's folder, commit, then deploy.
- Build Setup → Install, then merge the build workflow pull request: Do It For Me (Owner or Admin, after you tick the confirmation) merges it for you, or I'll Do It Myself means reviewing and merging it on GitHub. That merge starts your first deploy.
- If a deploy fails because a setup step is missing, the failure shows a button that takes you straight to that step.
Every deploy is scanned and checked like any other. Roll Back To This on an earlier deploy puts that version back live.
Not yet: pull request previews on Cloudflare (a preview says so and deploys nothing; merge to deploy), other frameworks such as Astro, SvelteKit or Nuxt, and custom domains on a Worker. Fastify cannot run on Workers at all (its router builds code at runtime, which Workers do not allow), so deploy a Fastify app to Render, Railway, Vercel or your own server.
Walkthrough
- Go to Settings → Deploy Targets.
- Find the platform (e.g. Vercel) and click Connect. Follow the step-by-step guide on the card to create a token in that provider and paste it in. The account stays yours.
- Click Verify to confirm ZipLoom can reach the account.
- Bind a target to a project when you create or configure it.
- To remove one, click Disconnect — existing live deployments are unaffected; new deploys to that target stop.
Options & Settings
- Connect / Verify / Disconnect per platform.
- AWS (BYOC) — bring-your-own-cloud AWS is configured separately under Settings → AWS.
Edge Cases & Gotchas
- Verify before your first deploy — an unverified/expired connection fails the deploy at the push step.
- Disconnecting a target used by a project will block that project's future deploys until re-connected or re-pointed.
- If a connection fails, the card explains the provider's own reason (for example an IP allow-list on Vultr, or a missing permission on AWS) and how to fix it.
- Contabo commits you to a month when a server is created, so ZipLoom never retries a create on your account by guessing.
Related
Deploying A Project · Server Placement · Custom Domains · App Catalog · AWS (BYOC) · Self-Host Export (Own)
Last updated: 2026-11-17 · docs v5.3.0
Deploy & Build
Deploying A Project
What it is. Running a build+deploy of a project to its connected target. The Deployments page (/deployments) shows "every deploy across your projects, newest first," and is where you trigger, watch, and manage them.
Why it matters. This is where the scan gate does its job: the security + compatibility checks run before the app is reachable, so risky AI-generated code is blocked, not shipped.
Walkthrough
- Open Deployments in the sidebar (or a project's detail page).
- Start a deploy with Redeploy — re-run this project (or the deploy action on a new project). ZipLoom pulls the latest commit, auto-detects the framework, and runs the pre-deploy scan.
- Watch status in the list. Each deployment row shows Project · Branch / commit · Platform · Duration and a status:
- In flight — building/scanning/deploying now.
- Ready — live and serving.
- Failed — stopped (build error or a blocking security finding).
- Cancelled — stopped by a user.
- On a Failed build, open it to see the cause — ZipLoom's AI Build Failure Diagnosis names the actual reason in plain English and gives the fix.
- On Ready, use the project's Open live URL to view it.
Options & Settings
- Status filter — All / Ready / In flight / Failed / Cancelled.
- Redeploy this project — re-run the latest commit.
- Delete deployment — remove a single deployment record.
- Clear failed — bulk-remove failed deployments to tidy the list.
Edge Cases & Gotchas
- Critical security findings block the deploy by default — resolve or dismiss them (Security → Findings) to proceed.
- "No deployments yet" shows until a project's first deploy.
- A failed deploy leaves the previous Ready release serving (nothing is torn down on failure); combine with Auto-Rollback for regressions that pass build but break at runtime.
Related
Projects · Deploy Targets · AI Build Failure Diagnosis · Security · Auto-Rollback / Staged Rollout (Operate)
Last updated: 2026-09-22 · docs v4.6
Deploy & Build
AI Builder / AI Assistant
What it is. A conversational builder (/ai, "AI assistant") that takes an idea in plain language and moves it toward a deployable project. Prompt: "Describe what you want to build…".
Why it matters. It's the front door for starting something new without hand-wiring a repo — describe it, refine it, and compose it into a deploy.
Walkthrough
- Open AI builder in the sidebar.
- In the composer, describe what you want to build.
- ZipLoom ideates a plan/approach. From there use the action buttons:
- ▶️ Continue — proceed with the current direction.
- ✏️ Refine idea — adjust scope/requirements and regenerate.
- 📋 Use a template — start from a matching template instead.
- Compose From Your Repos — jump to the Stack Composer above the chat. An idea has no repositories yet, so composing starts from repositories you give it.
- In the Stack Composer, give it your repositories and Compose. Edit the plan (name, repository, branch, host and wiring for each service), then Create Projects: one project per service. Deploy In Order lists them dependencies first; each Deploy goes through the normal Projects → Deployments path (scan gate included).
Options & Settings
- Iterative chat: refine as many times as needed before composing.
- Template hand-off connects to the Templates catalog.
Edge Cases & Gotchas
- Output is AI-generated — it still passes the same security scan on deploy; nothing skips the gate because it came from the builder.
- The Stack Composer makes one project per service; it does not merge repositories into one deployment. It wires service URLs only (never keys or secrets), and a service whose linked service is not deployed yet deploys without that variable and says so in its log. Discover's multi-repo "Compose selected" is a separate entry point (see Discover's "coming soon" note).
Related
Templates · Discover · Deploying a Project · Security
Last updated: 2026-09-22 · docs v4.6
Deploy & Build
Templates
What it is. A catalog of starting points (/templates) you can launch straight into a new project.
Why it matters. It is the fastest way to a working, deployable app: skip the scaffolding and go straight into the same checked deploy flow as everything else.
Walkthrough
- Open Templates in the sidebar.
- Type in Search templates… to filter the catalog.
- Choose a template's action button to launch it.
- ZipLoom creates a project from the template's repository and takes you into the deploy flow.
Options & Settings
- Keyword search. Commonly used templates surface first.
Edge Cases & Gotchas
- A launched template becomes an ordinary project and deploys through the same pre-deploy checks as your own code.
- "No templates match your search. Try a broader term." means nothing matched; clear the search to see everything.
- Looking for a complete self-hosted app such as a CMS or analytics tool? See the App Catalog.
Related
App Catalog · Discover · Projects · AI Builder
Last updated: 2026-11-10 · docs v5.1
Deploy & Build
Discover
What it is. Open-source discovery (/discover): "Search open source, then launch it live in one click." Results show each repository's stars, forks and a quality signal, and ready-to-deploy apps that match your search.
Why it matters. Many apps start from an existing open-source project. Discover finds one and deploys it through ZipLoom's checks without leaving the platform.
Walkthrough
- Open Discover and Search repos… for a project.
- Read Open-Source Repos On GitHub and, where your search matches one, Ready-To-Deploy Apps For This Search from the App Catalog.
- Launch a repository to create a project from it and start the deploy flow.
- To deploy onto your own server: ZipLoom copies the repository into your GitHub account (a fork), you approve one setup change, and it deploys to your server. You may be asked to give the ZipLoom GitHub App access to the copy, which is one step in GitHub.
- To combine several repositories: select them, then choose Compose A Stack. The Stack Composer opens with those repositories filled in, and proposes how they fit together.
Options & Settings
- Keyword search. The Stack Composer takes up to 10 repositories; extra selections are left off the link.
Edge Cases & Gotchas
- A launched repository is third-party code. The pre-deploy checks run on it, so read the findings before you ship.
- GitHub turns Actions off in a fork. If the build does not start, open the copy's Actions tab and enable workflows; the build runs there.
- Nothing is created if the repository cannot run on the server you chose.
- "No repos found. Try a different search." means the search matched nothing.
Related
AI Builder (Stack Composer) · App Catalog · Templates · Server Placement
Last updated: 2026-11-10 · docs v5.1
Deploy & Build
Multi-Platform Deployment
What it is. One project, eleven deploy targets. ZipLoom detects your framework, builds the app, runs the security gate, and ships it to the platform you choose — Vercel, Netlify, Cloudflare Pages, Railway, Render, Google Cloud Run, AWS, DigitalOcean, Vultr, Contabo or Amazon Lightsail. The platform accounts stay yours; ZipLoom holds the credentials in the encrypted Secrets Vault and deploys on your behalf.
Why it matters. Most failed launches aren't code failures, they're configuration failures: the wrong build command, a missing environment variable, a runtime mismatch. Handling that once, centrally, means moving between platforms is a redeploy rather than a rewrite — and it's why leaving ZipLoom doesn't strand your app, because it was never hosted here in the first place.
Walkthrough
- Connect a platform under Settings → Deploy Targets (see Deploy Targets for the per-platform credential each one needs).
- Open a project and choose Deploy. Pick the connected target.
- ZipLoom auto-detects the framework and derives the build command, output directory, and runtime. Override any of it if your project is unusual.
- Environment variables are pulled from the Secrets Vault at deploy time — decrypted only for that deploy, never stored in plaintext.
- The security gate runs. Untriaged critical findings stop the deploy; other findings are reported and the release proceeds.
- On success you get a live URL. The deployment record keeps the logs, the findings, and the commit it shipped.
Options & Settings
- Target — which connected platform receives this deploy.
- Build overrides — build command, output directory, runtime version, when auto-detection guesses wrong.
- Environment variables — per project and per environment, resolved from the vault.
- Custom domain — attach your own hostname (see Custom Domains); certificates are issued and then monitored.
- Staged / canary rollout — release gradually instead of cutting over at once (see Operate).
Edge Cases & Gotchas
- Platforms differ in what they can host. A static site, a server-rendered app, and a long-running worker are not interchangeable; the composer picks a sensible target but a Node worker won't run on a static host.
- Credentials are per platform. Each target needs its own token or key, and a revoked token fails the deploy rather than silently falling back.
- A gated deploy is a stopped deploy, not a broken one. Fix the finding (or disposition it deliberately) and redeploy.
- Deployed apps stay live if you cancel. They run in your accounts. You lose scanning, monitoring, and self-healing — not the app.
Related
Deploy Targets · Deploying a Project · Secrets Vault (AES-256-GCM) · Security Scanning (Pre-Deploy) · Staged / Canary Rollout · Self-Host Export (No Lock-In)
Last updated: 2026-11-09 · docs v5.0
Deploy & Build
App Catalog
What it is. A catalog of trusted self-hosted apps that ZipLoom deploys from their official images onto a server you own, in App catalog in the sidebar.
Why it matters. Open-source tools like analytics, forms or automation can replace monthly subscriptions, but installing them safely takes work. The catalog does the setup, generates the passwords, and keeps your data on your server.
Walkthrough
- Open App catalog.
- If you have no server provider connected, choose Connect A Server Provider first.
- Pick an app and read what it does for your business.
- Choose your server provider and deploy. ZipLoom pulls the official image, sets up storage for its data, creates its secrets in your vault and puts it behind HTTPS.
Options & Settings
- Apps that need their own database get one alongside them on the same server.
- Each app's image is checked for known vulnerabilities on the server; the result is recorded as a deploy check.
- Each app's data lives on its own volume and stays on the server it was created on.
Edge Cases & Gotchas
- Catalog apps run on a server (VPS) you own, not on managed hosts like Vercel.
- An app built from an image has no repository, so source-code checks are not recorded for it. ZipLoom says so rather than showing them as passed.
- The catalog is new; its public pages are at
ziploom.dev/self-host.
Related
Server Placement · Deploy Targets · Secrets Vault (Secure)
Last updated: 2026-11-09 · docs v5.0
Deploy & Build
Server Placement
What it is. For apps that deploy to a server (VPS), the Where This Deploys card lets you choose between a fresh server and one you already pay for.
Why it matters. A new server for every small app adds up. Reusing one you already have costs nothing extra; a fresh one keeps apps apart.
Walkthrough
- Open a VPS project's page and find Where This Deploys.
- Choose a fresh server or an existing one. The card shows what each choice costs.
- Deploy. The choice also appears in the deploy log.
Options & Settings
- Nothing is created or charged until you deploy.
- Every app on a server listens only on the server itself and is reached through its web server over HTTPS, never on an open port.
Edge Cases & Gotchas
- Only an Owner or Admin of the workspace can change the placement.
- An app that keeps data never moves away from the server it lives on.
- Fresh servers are set up with key-only sign-in and automatic security updates.
Related
Deploy Targets · App Catalog · Custom Domains
Last updated: 2026-11-09 · docs v5.0
Deploy & Build
Custom Domains
What it is. Put your app on your own domain. ZipLoom shows the DNS records to enter at your registrar, checks when they take effect, and confirms HTTPS. On Vercel, Netlify, Cloudflare Pages, Render and Railway you choose: Do It For Me adds the domain to your hosting account for you, or I'll Do It Myself records it in ZipLoom and leaves your host untouched.
Why it matters. Your own domain looks professional, and getting DNS and certificates right by hand is a common way launches stall.
Walkthrough
- Open a project and go to the Domain & Build tab.
- Enter your domain. On a VPS, choose Save. On a managed host, choose Do It For Me (Owner or Admin, after you tick the confirmation) or open I'll Do It Myself and choose Save Without Adding, then add the domain on your host yourself.
- Follow How to enter this at your registrar under DNS & Certificate.
- Watch the status move from Waiting for DNS to Verified by your host, then HTTPS is reachable.
Options & Settings
- On managed hosts, "verified" is your host's own answer. If the host reports a problem or cannot be read, ZipLoom says so rather than showing a pass.
- On Render, adding domains beyond your Render plan may be billed by Render. Do It For Me's confirmation says so, every time.
- Custom domains are available on paid plans.
Edge Cases & Gotchas
- Points somewhere else means your DNS record targets a different address; correct it at your registrar.
- Not reachable over trusted HTTPS yet is normal until DNS points at the host and it has issued a certificate.
- Removing the domain in ZipLoom does not remove it from your hosting account.
- ZipLoom's own domains cannot be attached.
Related
Deploying A Project · Server Placement · Deploy Targets
Last updated: 2026-11-09 · docs v5.0
Secure
Security Scanning (Pre-Deploy)
What it is. Every project is scanned before it can deploy. The Security page (/security, "Pre-Deploy Security") runs five checks against a GitHub repo, returning findings graded Critical / High / Medium / Low / Info, each with What this means and How to fix.
Why it matters. AI-generated code routinely ships RLS gaps, leaked secrets, vulnerable dependencies, and misconfigured auth. The scan gate blocks a deploy on Critical findings — before the app is reachable.
The Five Checks (Tabs)
| Tab |
Checks |
Run it |
All-clear |
| Code |
Source scan — RLS gaps, secrets, injection, auth misconfiguration, unverified payment webhooks, records fetched by id with no owner check, raw request bodies saved straight to the database, unsafe deserialization, risky GitHub workflow settings |
Paste GitHub URL → Scan Code |
"No security issues detected…" |
| CVEs |
Every installed package version (from your lockfile, including indirect ones) against the OSV vulnerability database, plus known malicious packages. An unreadable lockfile is reported, never treated as clean |
→ Audit Dependencies |
"No known CVEs…" |
| Health |
Repo health score |
→ Score Health |
— |
| Licenses |
Restricted/copyleft (GPL, AGPL) |
→ Scan Licenses |
"No restricted or review licenses found." |
| Env Parity |
Drift across .env, .env.staging, .env.production |
→ Check Env Parity |
"All environment files are in parity." |
Walkthrough
- Open Security. 2. Pick a tab. 3. Paste a GitHub URL, click the action. 4. Read findings by severity; each shows the risk and the fix. 5. Triage with Finding Memory (Findings page) so it won't re-flag.
Options & Settings
Severity summary (Critical/High/Medium/Low/Info) across the top; per-tab actions listed above.
- Dependency findings show the published CVSS score and who scored it (see CVSS, CWE & NIST 800-53 Mapping).
- The same checks run automatically before every deploy and preview, on nightly schedules from LoomWorks up, and from CI with the GitHub Action.
Edge Cases & Gotchas
- ZipLoom needs repo access (connect GitHub first).
- Untriaged critical findings block the deploy by default.
- The code scan judges code, not text: words in comments, strings or documentation do not raise findings, and test, sample and documentation files are not treated as app code (secrets are still checked everywhere).
- Env Parity flags a var present in one file but missing in another.
Related
Findings & Finding Memory · Safe To Launch? · Attack Chains · Live Supabase Lockdown Check · Secrets Detection · Dependency & License Audit · Auth & RLS · CI, API Keys & The PR Check
Last updated: 2026-11-09 · docs v5.0
Secure
Findings & Finding Memory
What it is. The inbox (/security/findings) collecting every finding across scans, and the Finding Memory system that lets you triage each once.
Why it matters. Without memory, the same issue re-flags every scan. Triage once and the Brain remembers your decision.
Walkthrough
- Open Findings. 2. Review by severity (Critical/High/Medium/Low/Info). Each open finding also carries a priority word (Low, Guarded, Elevated or High) that orders the list by how likely it is to be targeted. 3. For each, choose a disposition: accept, dismiss, or fix. 4. Once dispositioned, the Brain stops surfacing it on future scans unless it changes/recurs.
Options & Settings
- Dispositions: Accept / Dismiss / Fix.
- Fix It For Me opens a pull request with the fix and confirms it worked.
- Copy Fix For Your AI Tool copies a paste-ready brief for your coding assistant.
- Attack Chains at the top show findings that combine into one path an attacker could walk.
- Set Aside By AI Review lists medium, low and info findings that two AI models agreed are not real problems; Keep In Problems brings one back.
- Recurrence Radar surfaces what keeps coming back or regressed after a fix (see Operate).
Edge Cases & Gotchas
- A dismissed finding can reappear if the underlying code changes (that's Recurrence Radar working).
- Critical findings still gate deploys until resolved or explicitly dispositioned.
Related
Security Scanning · Fix It For Me · Attack Chains · AI Review Of Findings · App Risk Level & Hardening Plan · Recurrence Radar (Operate) · Governed Safety Rules (Own)
Last updated: 2026-11-09 · docs v5.0
Secure
Secrets Vault (AES-256-GCM)
What it is. Encrypted storage for your environment variables: encrypted at rest with AES-256-GCM, decrypted only for the deploy that needs them, and masked everywhere else.
Why it matters. Keys in a committed .env file, or in a dashboard that shows them back in plain text, stay readable by everyone who ever had access. The vault keeps them encrypted, masked and logged.
Walkthrough
- Open Secrets, select the project, and add each variable: its name (for example
DATABASE_URL), the value, and an optional note on what it is used for.
- Each value is encrypted with its own IV and authentication tag before it is written, so tampering is detectable. The plaintext is not stored.
- At deploy time the values are decrypted and injected into the build as ordinary environment variables.
- They are masked in build logs and in the interface, so a screen-share or a log does not leak them.
- To rotate a key, generate the new one at the provider, update the value here and redeploy.
Options & Settings
- Per-project variables. Flag a value for rotation as a reminder.
- Access follows your organization and roles, and every read and write is in the audit log.
Edge Cases & Gotchas
- The vault cannot recall a key that already leaked. If a value was ever committed or shared, rotate it at the provider.
- A secret committed to the repository is caught by Secrets Detection, not by the vault. Move it here and rotate it.
- Self-Host Export lists your variable names, never their values, so you know what to set elsewhere.
Related
Secrets Detection · Deploying A Project · Auth Audit & Lockdown Check
Last updated: 2026-11-10 · docs v5.1
Secure
Secrets Detection
What it is. A sweep of every file on every push that flags API keys, tokens and connection strings left in your code.
Why it matters. AI coding tools paste keys wherever it is convenient. Bots scan public repositories and deployed bundles continuously, and a found key is often used within minutes.
Walkthrough
- On every push, and in the code scan, ZipLoom reads every file and recent git history.
- It matches the published shapes of real provider keys, such as Stripe, OpenAI, AWS, Supabase service keys and GitHub tokens, and flags high-entropy strings that look like credentials even when they fit no known format.
- Each match becomes a finding with the file and line.
- Rotate the key at the provider first, then remove it from the code and store the new value in the Secrets Vault.
Options & Settings
- Values stored in the Secrets Vault are where they belong and are not flagged.
Edge Cases & Gotchas
- Rotate first, always. Removing a key from your files does not un-expose it. Check the provider's logs for activity you do not recognise.
- A key deleted in a later commit is still readable in git history by anyone who can clone the repository, so history is checked too.
- A key added to the wrong environment variable can end up in the client bundle shipped to every visitor. The live checks look for that as well.
Related
Secrets Vault · Security Scanning · GitHub Actions Integration
Last updated: 2026-11-10 · docs v5.1
Secure
Vulnerable Dependency & License Audit
What it is. Two dependency checks: CVE audit (every installed package version from your lockfile, including indirect ones, checked against the OSV vulnerability database, plus known malicious packages) and License Compliance Guard (flags GPL, AGPL, and other copyleft before they can force your source open).
Why it matters. AI-picked dependencies can be vulnerable or carry viral licenses. Both are caught pre-deploy.
Walkthrough
- In Security → CVEs, paste the repo → Audit Dependencies. Vulnerable packages list by severity (Critical/High/Medium/Unknown). 2. In Security → Licenses, → Scan Licenses; restricted/review licenses are flagged.
Edge Cases & Gotchas
- "No known CVEs…" / "No restricted or review licenses found." mean clean.
- New CVEs appear over time — re-audit periodically, not just at first deploy. Package Alerts tell you when a package you use is newly listed as exploited.
- Each finding shows the published CVSS score and who scored it.
- Licenses are read per package from your lockfile. AGPL, SSPL and restricted-use licenses block production deploys; GPL is advisory.
- An unreadable lockfile is reported as a partial check, never as clean.
Related
Security Scanning · Package Alerts · CVSS, CWE & NIST 800-53 Mapping · Change Impact (Operate) · Findings & Finding Memory
Last updated: 2026-11-09 · docs v5.0
Secure
Auth Audit & Lockdown (RLS) Check
What it is. Two configuration checks that run before the build: the Lockdown Check for row-level security (RLS), the missing-RLS class behind CVE-2025-48757, and the Auth Configuration Audit for token lifetimes, session rotation and CORS, the class behind the Moltbook exposure.
Why it matters. Both kinds of mistake leave the app working perfectly while data is exposed. A login that works proves the happy path; these checks look at what happens when a key or a token is in the wrong hands.
Walkthrough
- Run Security → Scan Code, or deploy. Both checks run before the build.
- The Lockdown Check reads your migration files on every deploy. Connect your Supabase project (see Live Supabase Lockdown Check) and production deploys also check the live database, table by table.
- The Auth Audit checks token lifetimes, whether sessions are renewed at sign-in and ended at sign-out, refresh-token rotation, CORS policy, and password and sign-up rules.
- Findings appear with What this means and How to fix. Fix and re-scan.
Options & Settings
- A table the public API exposes with RLS off, or a policy that lets signed-out visitors write to every row, stops the deploy.
- A policy that lets everyone read is flagged for you to confirm, since some tables really are public.
Edge Cases & Gotchas
- A table meant to be public: switch RLS on and add a read policy that says so. A table with RLS switched off entirely is treated as open and blocks the deploy.
- CORS with a wildcard origin and credentials lets any website call your API as your signed-in user. It is convenient in development and dangerous in production.
- ZipLoom never changes your policies or settings. It names the table or setting and the fix.
- The RLS check targets Postgres row-level security as Supabase uses it. Other databases are covered by the broader code and posture checks.
Related
Live Supabase Lockdown Check · Security Scanning · Findings · Governed Safety Rules
Last updated: 2026-11-10 · docs v5.1
Secure
GitHub Actions Integration & Compliance Mapping
What it is. Two ways security extends beyond the deploy: the ZipLoom Security Scan GitHub Action gates every merge in your CI pipeline, and Compliance Readiness Mapping maps SOC 2, HIPAA and GDPR controls to your deploy, with a PDF export that also maps findings to NIST SP 800-53.
Why it matters. Security belongs in the pipeline, not just at deploy — and buyers/auditors ask for compliance evidence.
Walkthrough
GitHub Actions: create an API key, add the action to your workflow and choose the severity that fails the job. Full steps and exit codes: CI, API Keys & The PR Check.
Compliance: open Compliance (/compliance), review the mapped controls (SOC 2 / HIPAA / GDPR) against your deploy, and export the readiness PDF. Each readiness label is Ready, Needs Review or Gaps Found.
Production deploys also pass a license gate on every plan (AGPL, SSPL and restricted-use licenses block; GPL is advisory) and, on Business, a compliance gate.
Edge Cases & Gotchas
- The Action mirrors the platform's scan rules (Governed Safety Rules), so pipeline and deploy stay consistent.
- Compliance mapping is a readiness aid, not a certification. The NIST 800-53 section is a mapping only, not an assessment, authorization or certification.
- Exporting the PDF or the evidence ledger needs an Owner or Admin, on Business.
Related
CI, API Keys & The PR Check · CVSS, CWE & NIST 800-53 Mapping · Evidence Ledger & Audit Chain (Own) · Security Scanning
Last updated: 2026-11-09 · docs v5.0
Secure
Safe To Launch?
What it is. A card at the top of every project page that answers one question: is this app ready to go live? It shows Yes, Not Yet or Not Checked Yet, and lists up to three fixes that matter most.
Why it matters. A long findings list does not tell you whether to ship. This card does, and it puts the most urgent fixes first so you know where to start.
Walkthrough
- Open Projects and choose a project.
- Read the Safe To Launch? card. It shows the verdict and the reason for it.
- Under Fix These First, work through the top fixes. Lower-priority items appear under Worth Fixing When You Can.
- On any fix, choose Copy Fix For Your AI Tool. It copies a ready-to-paste brief (what is wrong, where, and how to fix it) for Cursor, Claude Code, Copilot or any other assistant.
- Fix, push, and the next scan updates the verdict.
Options & Settings
- Fix order. Fixes are ordered by how likely each problem is to be targeted, using public records of which weaknesses attackers use. With no such data, they fall back to severity order.
- Public-data notes. A dependency fix can carry one short note, for example that the flaw is on CISA's list of vulnerabilities exploited in the wild. Notes appear only when the public data exists.
- Unmaintained building blocks. If a package named in a dependency finding shows no maintainer activity (OpenSSF Scorecard), the card says so and names the source.
Edge Cases & Gotchas
- Yes needs a complete scan of your code under 30 days old and no open critical or high findings. Findings you triaged as an accepted risk or a false alarm are not counted, and the card says how many.
- Not Yet means open critical or high findings, or a latest deploy stopped by a check.
- Not Checked Yet means no complete, recent scan exists, or the scan could not read all of your code. Link your GitHub repository to the project first.
- A yes is not a guarantee: ZipLoom's checks do not catch everything.
- The verdict reflects the latest completed scan. A scan that failed or never started is never shown as a pass.
- The same top fixes appear in the Monday summary email.
Related
Fix It For Me · App Risk Level & Hardening Plan · Findings & Finding Memory · Weekly Summary Email
Last updated: 2026-11-09 · docs v5.0
Secure
Fix It For Me
What it is. One button on a finding that asks ZipLoom to write the fix, open it as a pull request on your repository, and then scan the pull request to confirm the problem is gone.
Why it matters. Knowing what is wrong is half the job. Fix It For Me drafts the change, proves it worked, and still leaves the final decision to you, because ZipLoom never merges anything.
Walkthrough
- Open Security → Findings, or a project's Safe To Launch? card.
- On a finding that offers it, choose Fix It For Me.
- If your workspace requires it, an Owner or Admin chooses Approve And Open PR.
- ZipLoom opens a pull request and shows Checking the fix… while it scans the pull request's code.
- The result appears on the finding:
- Fix confirmed — merge the pull request: the problem is gone in the pull request.
- The problem is still there: the change did not remove it. Review the pull request or fix it by hand.
- Could Not Check: the verification scan could not run. Nothing is assumed fixed.
- Merge the pull request yourself on GitHub. When ZipLoom sees the merge it shows Fix confirmed and merged, and you can choose Deploy Now.
Options & Settings
- Every deploy check still runs on the merged code before it goes live.
- If the pull request is closed without merging, the finding stays open.
Edge Cases & Gotchas
- Some findings have no automated fix; the button then says so.
- Fix It For Me is available on every plan to members who can edit. Checking a merged fix uses one scan from your allowance.
- ZipLoom never pushes to your main branch; the fix is always a branch and a pull request, and some changes are refused rather than proposed.
- The fix is written by a paid AI model. If no suitable model is available, ZipLoom refuses rather than using a weaker one.
- ZipLoom needs write access to the repository through its GitHub App to open the pull request.
Related
Safe To Launch? · Findings & Finding Memory · Security Scanning
Last updated: 2026-11-09 · docs v5.0
Secure
Attack Chains
What it is. A panel that links separate findings into one path an attacker could actually walk, such as a leaked key plus a database table that key can read. Each chain is ranked above the same findings on their own, with the one step to fix that breaks it.
Why it matters. Three medium findings can add up to one critical problem. Fixing the right link first is faster than working down a severity list.
Walkthrough
- Open Security → Findings and find the Attack Chains panel at the top.
- Read each chain from the first step to the last.
- Fix the marked step. Once it is fixed, the chain breaks.
- Rescan to confirm the chain is gone.
Options & Settings
- Chains combine code findings with what the live site exposes, including what the post-deploy crawl finds once an app is live.
- A chain can include a finding you previously accepted or suppressed. The panel flags this, because a decision made about one finding on its own may not hold when it is part of a chain.
Edge Cases & Gotchas
- Chains only appear when the findings involved are present together; a single finding is not a chain.
- Ask your coding agent over MCP to "fix whatever the attack chains say to fix first".
Related
Findings & Finding Memory · Live Posture Scan (Operate) · MCP For Coding Agents (Account)
Last updated: 2026-11-09 · docs v5.0
Secure
App Risk Level & Hardening Plan
What it is. Two cards on each project page for workspace Owners and Admins. App Risk Level gives one word (Low, Guarded, Elevated or High) for how exposed the app is right now. Hardening Plan groups its open problems into Do First, Do Next and Do Later. Each finding in the findings list also carries a priority word.
Why it matters. It turns a scan into an order of work, based on how likely attackers are to target what was found, not only on how severe it looks.
Walkthrough
- Open a project. The two cards appear after the first full scan of your code.
- Read App Risk Level and the sentence under it.
- Work through Hardening Plan from Do First. Each group says where to start.
- Optional: in the project's Settings tab, answer the two Risk Profile questions: does this app handle personal, payment or other sensitive data, and do people sign in to it?
Options & Settings
- Priority words also appear on each open finding in Security → Findings, in the scan API and in the MCP tools, so your CI and coding agents see the same order.
- Risk Profile answers help ZipLoom put the fixes that matter most first. "I don't know" is the default and is never treated as "no". Changes are picked up within a day.
- The risk level is raised by one step if the app was recently rolled back after a failure, or failed a health check.
Edge Cases & Gotchas
- ZipLoom shows words, never the underlying numbers.
- Not Available Yet means no full scan has run.
- Only Owners and Admins see the risk level and the plan, on every plan. Members still see the priority word on findings they can open.
- The weekly summary email adds a trend line ("better", "unchanged" or "more serious") once two comparable weeks exist.
Related
Safe To Launch? · Findings & Finding Memory · Weekly Summary Email (Operate)
Last updated: 2026-11-09 · docs v5.0
Secure
Live Supabase Lockdown Check
What it is. Connect a Supabase project, and on every deploy ZipLoom reads its live settings for tables with row-level security (RLS) off, policies that let everyone in, and views or functions that bypass your policies.
Why it matters. Your code can be perfect while the database is wide open. A table exposed through the public API with RLS off is the most common way AI-built Supabase apps leak data.
Walkthrough
- Open Settings and find the Database Connection card.
- Enter the Supabase project reference or URL, and a Supabase personal access token (it starts with
sbp_).
- Choose Run Check to see the result immediately.
- From then on, the check runs on every deploy.
Options & Settings
- Blocks Deploys. A table that is exposed through the public API with RLS off stops the deploy. Other problems are reported as findings.
- Remove the connection at any time from the same card.
Edge Cases & Gotchas
- A Supabase personal access token can reach every project your Supabase account can see. ZipLoom only calls read endpoints and stores the token encrypted, but Supabase cannot limit a token to one project. Use a dedicated Supabase account or token, and revoke it if you stop using the check.
- If Supabase cannot be read, the project is reported as not checked, never as safe. If you revoked the token, remove the connection and connect again.
- If only part of the project could be read, the result says it is not a full check.
Related
Auth & RLS (Secure) · Deploying A Project (Deploy & Build)
Last updated: 2026-11-09 · docs v5.0
Secure
Private Pages Check & Decoy Rows
What it is. Two tripwires on a project's Protection tab. Private Pages Check tests up to ten paths you list (such as /admin) every day while signed out. Decoy Rows gives you a fake credentials row to plant in your own database, with a link only an intruder would follow.
Why it matters. Scans find mistakes in code. These tell you when something is actually exposed, or actually being used, in production.
Walkthrough: Private Pages Check
- Open a project, go to the Protection tab and find Private Pages Check.
- Enter your private paths, one per line, and save. The first check runs with the next daily re-check.
- Each path shows Reachable Without Signing In, Refused Or Redirected or Could Not Be Checked.
- A reachable path becomes a high-severity finding and sends an alert.
Walkthrough: Decoy Rows
- Open a project, go to the Protection tab and find Decoy Rows. Give the decoy a label, such as "Production database".
- Copy the SQL straight away; it is not shown again. Run it in your own database.
- If anyone opens the link stored in that row, ZipLoom alerts your alert destinations.
Options & Settings
- Alerts go to destinations in Settings → Alerts that subscribe to "Private page reachable without signing in" or "Decoy row link opened".
- Revoke a decoy to stop its alerts; then drop the row from your database.
Edge Cases & Gotchas
- Only Owners and Admins can edit the paths or create decoys.
- A decoy alert fires when the link is used, not when the row is read. A plain read cannot be detected.
- ZipLoom never connects to your database for decoys; you run the SQL yourself.
- An app that returns its normal page for every path can make a private path look reachable; check the result before acting.
Related
Alerts (Operate) · Live Posture Scan (Operate) · Findings & Finding Memory
Last updated: 2026-11-09 · docs v5.0
Secure
Security Advisories & Incident Text Messages
What it is. When ZipLoom publishes a security advisory about a widely used package or weakness, it appears in Security → Advisories with what to do and whether it matches anything in your workspace. Affected workspaces are re-scanned and notified. You can also opt in to a text message for incidents that affect apps like yours.
Why it matters. When a serious flaw hits a popular package, the first question is "am I affected?". Advisories answer that for your own apps instead of leaving you to read every alert on the internet.
Walkthrough
- Open Security → Advisories to see published advisories.
- Open one to read what happened, what to do, and whether it matches this workspace.
- If it matches, your affected projects are queued for a re-scan. The new findings appear in Security → Findings.
- Optional: in Settings, find Text Messages, add your mobile number, choose Text Me A Code, and confirm the code.
- Tick "Text me when a security incident affects apps like mine". A separate box covers product news; it is never ticked for you.
Options & Settings
- Every advisory is reviewed and approved by a person at ZipLoom before anything is sent.
- Advisory re-scans never count against your monthly scan allowance.
- A public list at
ziploom.dev/advisories shows each advisory's id, date and scope only.
Edge Cases & Gotchas
- Text messages are optional. If the card says texts are not available yet, email alerts are unaffected.
- Reply STOP to any text to turn texts off. Message and data rates may apply.
- Changing your number turns both options off; tick them again for the new number.
- Consent records are kept for up to four years. Before an incident-text consent would lapse, ZipLoom reminds you by email (and by text while the consent is active), and Settings shows a one-click Confirm for the final 90 days.
Related
Package Alerts · Findings & Finding Memory · Alerts (Operate)
Last updated: 2026-11-09 · docs v5.0
Secure
Package Alerts
What it is. An alert when a package your app uses is newly listed by CISA as exploited in the wild, or when public forecasts of attacks on it rise sharply within a week.
Why it matters. A dependency that was a low priority yesterday can become urgent overnight. Package alerts tell you when the outside world changes, without waiting for your next scan.
Walkthrough
- Nothing to set up for email. Workspace Owners and Admins receive one email per flaw and reason.
- A banner on the dashboard links to the affected findings. Choose Open Your Findings.
- To also send these to Slack or a webhook, open Settings → Alerts and subscribe a destination to "A package you use was newly listed as exploited, or its attack forecasts rose".
- Update the package, push, and rescan.
Options & Settings
- Alerts cover open dependency findings from your latest scan.
- On a white-label workspace, emails carry the agency's brand.
Edge Cases & Gotchas
- One alert per flaw and reason: the same news does not arrive twice.
- The wording repeats what the public sources say. ZipLoom does not claim an attack on your app.
Related
Dependency & License Audit · Security Advisories · Alerts (Operate)
Last updated: 2026-11-09 · docs v5.0
Secure
AI Review Of Findings
What it is. A second opinion on new findings from two AI models, which judge whether each one is a real problem in context. When both agree a medium, low or info finding is not real, it is set aside, and you can bring it back with one click.
Why it matters. Rule-based scanners raise some false alarms. Setting obvious ones aside keeps your list to problems worth your time, without hiding anything serious.
Walkthrough
- Open Security → Findings.
- If any were set aside, an expandable section reads N Findings Set Aside By AI Review, with the reason for each.
- If you disagree, choose Keep In Problems. The finding returns to your list and stays there.
Options & Settings
- A finding is set aside only when both models agree. Anything unclear is kept.
- Critical and high findings are never set aside. A person always decides those.
- Only findings that nobody has triaged yet are reviewed.
Edge Cases & Gotchas
- The models see a short, masked piece of the surrounding code and the finding's description. No code is stored.
- Your workspace may have AI review turned off; then nothing is set aside.
Related
Findings & Finding Memory · Security Scanning
Last updated: 2026-11-09 · docs v5.0
Secure
CVSS, CWE & NIST 800-53 Mapping
What it is. Standard reference data next to your findings. Dependency findings show the CVSS score published for the vulnerability, with who scored it. Findings carry their CWE weakness type. The compliance readiness report adds a mapping of findings to NIST SP 800-53 controls.
Why it matters. Security teams, auditors and customers speak in CVSS, CWE and 800-53. Having them beside each finding saves you looking them up and makes reports easier to hand over.
Walkthrough
- Open Security → Findings or a deploy's dependency audit. A dependency finding shows a line such as "CVSS 3.1: 9.1 Critical — scored by NVD."
- Open a finding to see its CWE.
- On Compliance, an Owner or Admin exports the readiness PDF. It includes a section mapping findings to SP 800-53 Rev. 5 controls.
Options & Settings
- CVSS is shown exactly as published: NVD's own score first, otherwise the score from the organization that reported the flaw, named as such. ZipLoom never computes its own score.
- No published score means no line, never a zero.
- CVSS does not change a finding's severity or its place in your fix order.
- For a dependency finding, the CWE is the one NVD records for that vulnerability.
Edge Cases & Gotchas
- The 800-53 section is a mapping only. It is not an assessment, authorization or certification, and ZipLoom is not affiliated with or endorsed by NIST.
- Some findings have no fitting control; the report counts them as unmapped rather than leaving them out.
- NVD-derived data carries NVD's required notice wherever it appears.
- Exporting the readiness PDF is available on Business and needs an Owner or Admin.
Related
GitHub Actions & Compliance · Dependency & License Audit · Findings & Finding Memory
Last updated: 2026-11-09 · docs v5.0
Secure
CI, API Keys & The PR Check
What it is. Run ZipLoom's source scan from your own CI with an organization API key and the ZipLoom GitHub Action, and see a ZipLoom Security check on pull requests that have a preview deploy.
Why it matters. It stops a vulnerable change at the pull request, before it is merged, and in the tools your team already uses.
Walkthrough
- In Settings, find API Keys. Name the key (for example "GitHub Actions — main repo"), choose CI / GitHub Action, and create it.
- Copy the key immediately and store it as a GitHub secret named
ZIPLOOM_API_KEY. Choose I Have Stored It.
- Add the ZipLoom Security Scan action to your workflow with your
api-key and project-id.
- Set
fail-on to the lowest severity that should fail the job (default high).
Options & Settings
- Exit codes: 0 passed · 1 findings at or above
fail-on · 2 scan error (failed, timed out, or scanned a different commit, never a pass) · 3 configuration error (bad input, revoked key, unknown project).
- The action writes
ziploom-scan.json and, for a completed scan, ziploom-scan.sarif for GitHub code scanning.
- PR check: on pull requests with a preview, ZipLoom posts a ZipLoom Security check and one summary comment, kept up to date: Blocked or Passed, with a reference number.
Edge Cases & Gotchas
- Only Owners and Admins can create or revoke keys. A revoked key stops working at once.
- A key can scan and read results only for its own organization's projects.
- The PR check never lists finding counts or details; open ZipLoom for those.
- Making the check required for merging is a setting in your GitHub repository.
Related
MCP For Coding Agents (Account) · Security Scanning · GitHub Actions & Compliance
Last updated: 2026-11-09 · docs v5.0
Operate
Staged / Canary Rollout
What it is. Gradual traffic shifting across rings — Canary → Early ring → Half traffic → Full deploy (1% → 10% → 50% → 100%) — with an automatic error gate between rings. Managed at /rollout. Business+ tier, and only for projects deployed to Google Cloud Run.
Why it matters. A bad release reaches 1% of users, not 100%. If the error gate trips, the rollout stops before wider exposure.
Walkthrough
- Turn staged rollout on for the project in
/rollout settings (Business+, Google Cloud Run projects only). There is no start button: once it is enabled, the next deploy of that project runs as a staged rollout, and /rollout shows the active one and its history.
- The rollout advances through rings; status shows Pending / In progress / Paused / Complete / Rolled back.
- Controls: advance (promote to the next ring), pause, resume, rollback.
- The error gate auto-pauses/rolls back a ring that breaches the threshold.
Options & Settings
- Rings: Canary, Early ring, Half traffic, Full deploy.
- Manual pause/resume/advance override the automatic progression.
Edge Cases & Gotchas
- Business+ feature, and available for Google Cloud Run projects only — a project on any other host is refused.
- A rollout is started by the deploy itself, not by a button — enable it for the project first, then deploy.
- "No active rollouts." means nothing is mid-rollout.
Related
Deploying a Project · Auto-Rollback · Perf Budget · Synthetic Monitoring
Last updated: 2026-09-22 · docs v4.6
Operate
Auto-Rollback
What it is. A watch on live deploys that automatically restores the last good release when a deploy regresses. Page: /rollback (Active Deploy Watches, Rollback History, Rollbacks Triggered). Available on every plan.
Why it matters. Catches regressions that pass the build but break at runtime — without you watching dashboards.
Walkthrough
- Open Auto-rollback. Active watches show status: Watching / Rolled back / Disabled / Expired.
- A watch monitors a live deploy against the Default Threshold.
- On breach, ZipLoom automatically restores the previous good deploy and records it in Rollback History. A watch starts only after a deploy that has an earlier good deploy behind it.
Options & Settings
- Default Threshold — the error level that triggers a rollback.
- Per-deploy policies (rollback.getPolicies).
Edge Cases & Gotchas
- Not plan-gated. Watches are armed automatically on every plan, and the page and its history are open to everyone. Registering a watch by hand, with your own thresholds, is Business+ and currently has no screen in the app — the automatic watch above is what every plan gets.
- Hosts. Automatic restore works on Cloudflare Pages, Vercel, Netlify, Render, Railway, Google Cloud Run and your own AWS. A self-managed VPS (Vultr, Contabo, DigitalOcean, Amazon Lightsail) gets a new server on every deploy, so it has no watch and no restore. A project's first deploy has nothing to restore either.
- A watch becomes Expired when a newer deploy replaces it; a Disabled watch won't act.
- Complements Self-Healing Deploys (uptime restore) and Staged Rollout (pre-100% gating).
Related
Self-Healing Deploys · Staged Rollout · Deploying a Project
Last updated: 2026-09-22 · docs v4.6
Operate
Self-Healing Deploys
What it is. A health check every 5 to 15 minutes, at your plan's cadence (every 15 minutes on Free, Solo and Guild; every 5 minutes on Studio, LoomWorks, Business and Agency); if the site goes dark, the last good deploy is restored automatically. Automatic restore covers the managed hosts — Cloudflare Pages, Vercel, Netlify, Render, Railway, Google Cloud Run and your own AWS. It does not cover a self-managed VPS (Vultr, Contabo, DigitalOcean, Amazon Lightsail): each VPS deploy builds a brand-new server, so there is no earlier deploy to restore. The deploy's timeline says so, and ZipLoom shuts the superseded server down so it stops billing.
Why it matters. Downtime gets fixed before you (or your users) notice, without manual intervention.
Walkthrough
- Once a project is live, ZipLoom probes it on your plan's cadence (every 5 to 15 minutes). 2. If a probe finds the site down, and the host supports restore, it automatically restores the last known-good deploy. 3. The action is logged.
Edge Cases & Gotchas
- This restores availability (site dark); Auto-Rollback handles regressions (error-rate breaches) — they're complementary.
Related
Auto-Rollback · Synthetic Monitoring · Live Posture Scan
Last updated: 2026-11-09 · docs v5.0
Operate
Synthetic Monitoring
What it is. Post-deploy smoke tests against your live site (/monitor): your home page, an optional health path and robots.txt, checked within seconds of each deploy, plus a Spot Check you can run on demand. LoomWorks and above.
Why it matters. A deploy can report success while the site returns errors. A smoke test right after the release catches that before your users do.
Walkthrough
- Open Monitor and finish any Setup required step.
- Optionally set a health path, such as
/healthz, that your app answers when it is healthy.
- Deploy. The checks run automatically afterwards.
- Run a Spot Check on one of your projects at any time.
- Each run reads Pass, Fail, Skipped or Running; open one for the detail. The page also shows the 30-day pass rate, runs and failed runs.
Options & Settings
- Optional health path per project; manual spot checks.
Edge Cases & Gotchas
- LoomWorks and above. On lower plans the page offers See plans.
- These checks run after deploys and on demand, not on a recurring timer. For continuous watching with automatic rollback, see Self-Healing Deploys.
- "No smoke test runs yet." and "No check results yet." until the first run. A run that could not complete reads as skipped or failed, never as a pass.
Related
Self-Healing Deploys · Live Posture Scan · Auto-Rollback
Last updated: 2026-11-10 · docs v5.1
Operate
Performance Budget (Lighthouse CI)
What it is. A Lighthouse audit (through Google PageSpeed Insights) after every production deploy (/perf), scored for Performance, Accessibility, Best Practices and SEO and checked against the budget you set. LoomWorks and above.
Why it matters. Slow pages do not throw errors. They lose users and search ranking quietly. A budget attaches the number to the release that changed it.
Walkthrough
- Deploy to production. The audit runs automatically afterwards.
- Open Perf budget and select a project to see its history.
- Each deploy reads Within this project's budget or Flagged for review on this deploy.
- Set the budget: a minimum Lighthouse score (0–100) per category, and optional maximums for page metrics such as load and layout shift. Leave a maximum empty to not budget it. Changes are saved and apply from the next deploy.
Options & Settings
- Per-project budgets; a default budget applies until you change it.
- A drop larger than your threshold since the previous deploy also flags the release.
Edge Cases & Gotchas
- LoomWorks and above.
- A flag is not a rollback. It marks the release for review; it never undoes the deploy.
- Audits run against the live URL after production deploys, not on previews or local builds.
- Pages behind a sign-in are not measured; the audit sees what a signed-out visitor sees.
- "No Lighthouse audits recorded yet." until the first production deploy.
Related
Deploying A Project · Change Impact · Synthetic Monitoring
Last updated: 2026-11-10 · docs v5.1
Operate
Change Impact
What it is. Analyse a dependency update before you take it (/impact): Semver Risk, Dependency Blast Radius, Migration Notes and a Changelog summary. LoomWorks and above.
Why it matters. A version bump can break a build or change behaviour quietly. Seeing what it touches before you upgrade turns a surprise into a planned change.
Walkthrough
- Open Change impact.
- Enter the package (for example
react), the version you are on (17.0.2) and the version you want (18.2.0), then choose Analyze Update.
- Read the report:
- Semver Risk — Patch ("bug fixes only"), Minor ("additive, usually safe") or Major ("breaking changes likely").
- Dependency Blast Radius — every package affected by the update.
- Migration Notes — what you will need to change.
- Changelog — what changed between the two versions.
- Earlier analyses stay under Recent Analyses.
Edge Cases & Gotchas
- LoomWorks and above. On lower plans the page offers See plans.
- "No analyses yet. Enter a package update above." until you run one.
- Semver risk describes what the version numbers promise. A package that breaks its own promise can still surprise you, so run your tests after upgrading.
Related
Vulnerable Dependency Audit · Deploying A Project · Performance Budget
Last updated: 2026-11-10 · docs v5.1
Operate
Runbooks
What it is. AI-generated operational runbooks for your projects (/runbooks): written procedures for recurring tasks and incidents, saved where the team can find them. LoomWorks and above.
Why it matters. The steps for restarting, rolling back or checking an app usually live in one person's head. A written runbook means anyone on the team can follow them at 2 a.m.
Walkthrough
- Open Runbooks and choose New Runbook.
- Enter the Project name (for example
my-saas-app) and, optionally, the App URL.
- Choose generate. While it works the button reads Generating….
- Select a runbook from the list to read it. Regenerate refreshes one after your app changes.
Options & Settings
- One runbook per project; regenerate whenever the app or its setup changes.
Edge Cases & Gotchas
- LoomWorks and above. On lower plans the page offers See plans.
- A runbook is a starting point written by AI from what it is told. Read it and correct anything specific to your setup before you rely on it.
- "No runbooks yet." and "Select a runbook to read it." are the empty states.
Related
Logs · Deploying A Project · Self-Healing Deploys
Last updated: 2026-11-10 · docs v5.1
Operate
Logs & Audit Events
What it is. The workspace event log (/logs): every deployment step and every action taken in your workspace, newest first. LoomWorks and above.
Why it matters. When something changes unexpectedly, the first question is who did what and when. The log answers it across deploys, settings and members, without anyone having to remember.
Walkthrough
- Open Logs.
- Narrow the list by actor (a person, or All actors) and by level: Info, Warn or Error.
- Read down the stream. Each entry names the actor, the action and the time.
- Deploy something to see each pipeline step appear as it happens.
Options & Settings
- Actor and level filters, combined.
- A separate organization audit log of settings changes sits under Settings.
Edge Cases & Gotchas
- LoomWorks and above. Lower plans see the plan needed instead of the list.
- "No events yet. Deploy something, and every step will appear here." on a new workspace.
- "No events match these filters." means the filters are too narrow, not that nothing happened.
- Sensitive actions are also written to the tamper-evident audit chain, which records field names and never secret values.
Related
Evidence Ledger & Audit Chain · Settings & Workspace · Deploying A Project
Last updated: 2026-11-10 · docs v5.1
Operate
Costs & AI Spend
What it is. Cost visibility for your workspace (/costs): Total This Period, Spend by Model, By Provider, daily spend for the last 30 days, recent requests, and a Budget Alert.
Why it matters. AI usage costs creep in small amounts that are invisible one call at a time. Seeing them by model and by day, with an alert you set, keeps them bounded.
Walkthrough
- Open Costs.
- Read the period total, then Spend by Model and By Provider to see where it goes.
- Check Daily Spend — Last 30 Days for spikes, and Recent Requests and Recent Events for what caused them.
- Set a Budget Alert: "Notify when monthly spend exceeds a threshold."
Options & Settings
- Budget Alert threshold. Setting it needs an Owner or Admin; every member can view the page.
- MRR Cost Ceiling shows AI spend against the ceiling for this billing cycle. As spend nears the ceiling, ZipLoom moves work to lower-cost models. Past it, free models only, until spend falls back. A banner says when either is happening.
Edge Cases & Gotchas
- The ceiling resets each billing cycle on its own.
- "No MRR this cycle, so there is no ceiling to measure AI spend against." appears for a workspace with no paid subscription this cycle.
- "No cost data yet" and "No AI calls this period" until usage accrues. A card that could not load says so; it never shows a zero in place of data it could not read.
Related
Change Impact · Deploying A Project · Billing & Plans
Last updated: 2026-11-10 · docs v5.1
Operate
In-App Debug Copilot
What it is. A conversational debugging assistant built into the deployment view. Paste an error log or describe what went wrong, and it answers with your project's context already loaded — the build output, the detected stack, and the deployment history.
Why it matters. The usual loop is: copy the log, paste it into a separate chat, strip anything sensitive, re-explain your stack, repeat. The copilot starts from the deployment it's attached to, which removes the busywork and the risk of pasting a key into a tool that shouldn't see it.
Walkthrough
- Open the failing deployment (Deployments → the run).
- Scroll to the debug assistant. Paste the error log or describe the issue in the composer ("Paste error log or describe the issue…").
- Read the diagnosis. It names the likely cause rather than restating the error.
- Use Copy response to hand the fix to the AI tool that wrote the code.
- Apply the change and redeploy. The next run's logs and findings are attached to the new deployment record.
Typical prompts it's built for:
- "My deploy failed — here's the error log."
- "My app builds locally but fails on the platform. Why?"
- "How do I add environment variables to my deployment?"
- "Why is my app returning 500 errors in production?"
Options & Settings
- Copy response — one-click copy, formatted to paste into a coding agent.
- Context — scoped to the deployment you opened; open a different run to ask about that one.
- Free-form input — a pasted log and a plain-language question both work.
Edge Cases & Gotchas
- It explains and suggests; it does not change your code. Applying the fix stays your decision.
- It answers from the context it has. A problem that lives outside the deployment — a third-party outage, a DNS change made elsewhere — may need information it can't see.
- Don't paste anything into any assistant you wouldn't want stored. That applies here as much as anywhere else.
- For a build that failed outright, the deployment record already carries the automatic failure diagnosis (see Deploying a Project); the copilot is for asking follow-up questions.
Related
Deploying a Project · Logs & Audit Events · Self-Healing Deploys · Findings & Finding Memory
Last updated: 2026-09-22 · docs v4.6
Operate
Alerts: Slack & Webhooks
What it is. Send ZipLoom's important events to a Slack channel or to your own signed webhook, chosen per event.
Why it matters. Problems get fixed faster when they land where your team already works, instead of waiting for someone to open a dashboard.
Walkthrough
- Open Settings and find Alerts.
- Choose the destination type: a Slack incoming webhook URL, or your own webhook URL.
- Name it (for example "#deploys") and tick the events it should receive.
- For a webhook, copy the signing secret shown once and choose I have saved it.
- Choose Send test to confirm it works.
Options & Settings
Events you can subscribe to:
- Deploy failed
- Deploy blocked by a security check
- Rolled back to the previous version
- Staged rollout reverted
- New critical findings
- Decoy row link opened
- Private page reachable without signing in
- A package you use was newly listed as exploited, or its attack forecasts rose
Webhook requests carry a ZipLoom-Signature: t=…,v1=… header so you can confirm each one came from ZipLoom.
Edge Cases & Gotchas
- Destinations are visible to organization members, not viewers.
- A destination that keeps failing is turned off; the card shows Turned off.
- Webhook URLs must be public addresses; private or internal network addresses are refused.
Related
Private Pages Check & Decoy Rows (Secure) · Package Alerts (Secure) · Auto-Rollback
Last updated: 2026-11-09 · docs v5.0
Operate
Weekly Summary Email
What it is. "Your App This Week", a Monday email to your workspace's Owners and Admins with a short recap of your apps and the Three Things To Fix This Week.
Why it matters. One email a week keeps security moving without anyone having to remember to log in.
Walkthrough
- Nothing to set up: Owners and Admins receive it every Monday.
- Read the three fixes. Each shows its severity, project, file location, a plain explanation and the first steps of the fix.
- Follow the single link to Security → Findings to act on them.
- To stop it, open Settings and turn off Weekly Summary Email.
Options & Settings
- The three fixes are the same ones the project page's Safe To Launch? card puts first.
- The email shows each app's risk level, and a trend line ("better", "unchanged" or "more serious") once two comparable weeks exist.
Edge Cases & Gotchas
- The email carries counts and plain explanations, never code.
- An app with nothing open is simply not listed among the fixes.
Related
Safe To Launch? (Secure) · App Risk Level & Hardening Plan (Secure)
Last updated: 2026-11-09 · docs v5.0
Own Your Stack
Self-Host Export (No Lock-In)
What it is. The Exit Kit: download a self-host bundle for any project, at any time, on every plan. It contains a Dockerfile, a docker-compose.yml, a Terraform configuration, an .env.template, a setup README and a SQL file of the project's records in ZipLoom.
Why it matters. Lock-in is usually discovered at the worst moment, when you want to leave or a procurement review asks what happens if the vendor goes away. The answer here is a file you can open today.
Walkthrough
- Open the project and find the Self-Host Export card.
- Choose Download export package. ZipLoom builds the bundle.
- Open the zip and follow
SELF_HOST_README.md.
- Fill the
.env.template with your values and run the app on your own infrastructure.
What Is In The Bundle
Dockerfile and docker-compose.yml to build and run the app.
terraform/ with a starting configuration for a cloud host.
.env.template listing every variable your app needs, without values.
data/ziploom-data.sql with the project's records in ZipLoom: its deploys, settings and configuration.
Edge Cases & Gotchas
- Secret values are never in the file. Copy them from wherever you keep them.
- The SQL file is not a copy of your app's own database. That stays with your database provider.
- Your deployed apps run on your own hosting accounts, so they keep running if you cancel. You lose ZipLoom's scanning, monitoring and self-healing, not the app.
- Try the export during your trial. It is meant to be inspected before you commit.
Related
Deploy Targets · AWS (BYOC) · Governed Safety Rules
Last updated: 2026-11-10 · docs v5.1
Own Your Stack
White-Label The Dashboard
What it is. Put your logo and accent colour on the dashboard your clients see, and serve it on your own dashboard address. Configured in Settings → Branding.
Why it matters. When you build apps for clients, the security report and the deploy notices are part of what you deliver. Your brand on them makes the work read as yours.
Walkthrough
- Open Settings and find Branding.
- Set your Logo URL and Accent colour (for example
#1F3A5F), then choose Save branding.
- Set your Dashboard address and choose Save address.
- Create client workspaces under your account. Each one inherits your branding; you do not set it per client.
Options & Settings
- Logo URL, accent colour, dashboard address.
- Only Owners and Admins can change branding.
Edge Cases & Gotchas
- White-label is part of the Agency plans, which are not open for self-serve purchase at the moment.
- On a white-label workspace, deploy, rollback, scan and invite emails and the PDF report carry your brand. ZipLoom's own support chat and community link are never shown to your clients.
- The sending address and page links are still on ZipLoom's domain today.
- Branding changes presentation only. The findings in a report are unchanged, because they are what makes it worth handing over.
Related
Settings & Workspace · Billing & Plans · Team & Roles
Last updated: 2026-11-10 · docs v5.1
Own Your Stack
Governed Safety Rules
What it is. The rules that judge your code are versioned files that change only through a reviewed pull request.
Why it matters. If scan results shift and nobody can say why, they stop being evidence. Because every rule change carries an author, a date and a diff, a change in what you are told can be traced to a specific release.
Walkthrough
- Every scan records which rule-pack version judged it.
- Each finding and gate decision carries an opaque reference, such as
FX-1A2B3C4D, that your team can trace in the console and in the tamper-evident audit log.
- When results change between two scans of the same code, the rule-pack version tells you whether a rule release caused it.
- The same rules drive the in-app scan, the deploy gate and the
ziploom/scan-action GitHub Action, so all three agree.
Edge Cases & Gotchas
- You cannot edit the rules from your workspace. Changes go through review, which is what makes the results defensible to a customer or an auditor.
- Rule internals stay private on purpose. A finding tells you what is wrong, where and how to fix it. Which pattern fired is not shown anywhere, because a scanner that reveals its patterns becomes a tool for tuning attacks around them.
- AI explains and helps fix findings. Whether something is a finding is decided by the inspectable rules.
Related
Security Scanning · GitHub Actions Integration · Evidence Ledger & Audit Chain
Last updated: 2026-11-10 · docs v5.1
Own Your Stack
Trust Badge & Share Security Proof
What it is. Two ways to show others your app is checked, both on a project's Protection tab. The Trust Badge is a public "Verified by ZipLoom" badge for your README or website. Share Security Proof creates a private link to a no-login page and PDF showing your recent deploys and the verdict of every gate that checked them.
Why it matters. Customers, investors and buyers ask whether your app is secure. These answer with evidence instead of a promise, without exposing your findings.
Walkthrough: Trust Badge
- Open a project, go to Protection and turn on Trust Badge.
- Copy the snippet into your README or site.
- Anyone clicking the badge sees its verification page.
Walkthrough: Share Security Proof
- On the same tab, an Owner or Admin chooses how long the link lasts (7, 30 or 90 days).
- Copy the link straight away; it is not shown again.
- Send it. Revoke it at any time.
Options & Settings
- The badge is shown only when the latest completed scan is under 30 days old and has no open critical or high findings. Accepted risks still count against it.
- On paid plans without nightly scans, ZipLoom re-scans a badge-enabled project once its latest scan is 20 days old, so the badge does not lapse. On Free, Owners and Admins get a reminder email instead.
- The proof page covers the last 20 production deploys and the badge state.
Edge Cases & Gotchas
- Neither the badge nor the proof page ever shows finding counts, titles or file paths.
- A badge reflects automated checks at one point in time. It is not a guarantee that an app has no vulnerabilities.
Related
Evidence Ledger & Audit Chain · Security Scanning (Secure)
Last updated: 2026-11-09 · docs v5.0
Own Your Stack
Evidence Ledger & Audit Chain
What it is. A record of every deploy and the security gate decision that let it through, signed and tied to a tamper-evident audit log. Export it from Compliance as a PDF and JSON for any date range.
Why it matters. Auditors want proof that checks ran on what actually shipped. The ledger shows it, and the audit chain shows the record has not been edited afterwards.
Walkthrough
- Open Compliance and find Deploy-Gate Evidence Ledger.
- Choose the From and To dates.
- Choose Export PDF + JSON. ZipLoom verifies the audit chain while it builds the file.
Options & Settings
- Each gate's verdict is one of passed, blocked, advisory or error. ZipLoom never records a pass for a check that did not fully run.
- The ledger lists exceptions plainly, such as a live deploy with no recorded gate, a gate that failed open, a dependency check with no lockfile, or a suppression with no reason or expiry.
- What each gate checks explains every gate in plain words.
Edge Cases & Gotchas
- The audit chain is computed by the database. If it is broken, the export says Audit chain broken.
- Exports need an Owner or Admin, on Business.
Related
GitHub Actions & Compliance (Secure) · Trust Badge & Share Security Proof
Last updated: 2026-11-09 · docs v5.0
Account
Settings & Workspace
What it is. Your account and organization preferences (/settings): profile, notification emails, branding, dashboard address, API keys, repository access, the organization audit log and account deletion.
Walkthrough
- Profile — your display name. Your email is "managed by your sign-in provider", so change it there.
- Weekly Summary Email — a Monday recap of your apps, sent to organization Owners and Admins.
- Scan Report Email — sent when a repository is scanned for the first time, or when a scheduled scan finds new critical or high findings. Owners and Admins only.
- Branding — logo and accent colour. On Agency plans they apply to every client workspace under yours (see White-Label The Dashboard).
- Dashboard address — the subdomain your dashboard is served on; choose Save address.
- API keys — keys for the CLI, the GitHub Action and coding agents. Each key is shown once.
- Repository access — which GitHub repositories ZipLoom can deploy.
- Audit log — "Recent actions in this organization."
- Account deletion — request it here. You get an email confirming the request, and ZipLoom follows up before any data is deleted.
Options & Settings
Sub-pages: Team, Billing, Deploy Targets, AWS (BYOC), MCP Servers, Alerts.
Edge Cases & Gotchas
- Only Owners and Admins can change branding and organization-level settings, create API keys, or receive the weekly and scan report emails.
- An API key is shown once. If you lose it, revoke it and create a new one.
Related
Team & Roles · Billing & Plans · Deploy Targets · AWS (BYOC) · MCP Servers · Two-Factor Authentication
Last updated: 2026-11-10 · docs v5.1
Account
Team & Roles
What it is. Manage who can use your organization (/settings/team): invite people by email, set their role, and remove them.
Why it matters. Deploy access is production access. Roles keep sensitive actions, such as billing, branding, API keys and approvals, with the people who should hold them.
Walkthrough
- Open Settings → Team.
- Enter the email of the person to invite (for example
teammate@company.com) and choose their role.
- ZipLoom emails them a link. They join once they accept it, signed in with that same address, whether they are new to ZipLoom or not. Until then they show as Waiting to accept.
- Change a member's role, or remove them, from the member list.
Options & Settings
- Owner — full control of the organization.
- Admin — manages settings, branding, members, API keys and approvals.
- Member — uses the workspace: projects, deploys and scans.
- Cancel invite withdraws an invitation that has not been accepted.
Edge Cases & Gotchas
- Only an Owner or Admin can invite, change roles or remove people. Members see "Ask an owner or admin to add the person you have in mind."
- An invitation link is single-use and expires after 7 days. Send a new one if it lapses.
- The owner role cannot be changed here, and the organization's owner cannot be removed.
- A removal can be undone from the notice that appears. Share links the person created stay revoked even if you bring them back.
- If a change fails you see "That change was not saved. Try again." Nothing is half-applied.
Related
Settings & Workspace · Billing & Plans · Two-Factor Authentication
Last updated: 2026-11-10 · docs v5.1
Account
Billing & Plans
What it is. Plan and billing management in Settings → Billing: your current plan, what you have used this month, and the Stripe billing portal. Pricing is a flat subscription with no usage meters and no per-seat charges, monthly or annual (annual saves 10%). Current prices are always on ziploom.dev/pricing.
Why it matters. You know what you will pay every month. Allowances grow with each plan, so upgrading buys more room, not surprise bills.
Walkthrough
- Open Settings → Billing.
- See Current Plan and Current Usage (deploys, AI compositions and scans this month).
- Choose a plan to check out, or open the portal to change your payment method, switch plans, apply a promo code or cancel.
Options & Settings
- Plans: Free, Solo, Guild, Studio, LoomWorks and Business. Agency and Agency Pro are not open for self-serve purchase at the moment.
- Packs: add-ons such as extra apps, scans or AI compositions stack on top of your monthly allowance.
- Business-only capabilities: compliance PDF export, staged rollouts and deploying to your own AWS account (BYOC). Scheduled nightly scans and performance budgets start at LoomWorks.
- Workspaces bill through the parent account, and usage is pooled across them.
Edge Cases & Gotchas
- Self-healing is never plan-gated. Every plan gets automatic restore of the last good deploy on supported hosts; only how often your site is checked differs (every 5 to 15 minutes).
- A scan that never started (for example, one that was refused before it ran) does not count against your allowance, and a pack credit used for it is returned.
- Refunds: purchases made from 8 November 2026 can be refunded within 14 days; earlier purchases keep their 30-day window. See
ziploom.dev/refund-policy.
- Usage allowances reset each month.
Related
Founding Member Plans · Team & Roles · Settings & Workspace
Last updated: 2026-11-09 · docs v5.0
Account
AWS (BYOC — Bring Your Own Cloud)
What it is. Connect your own AWS account (/settings/aws) so deploys run in infrastructure you own. Static sites deploy to S3 behind CloudFront in your account. Business plan and above.
Why it matters. Some teams need the app, its data and its bill to sit in their own cloud account, for procurement, compliance or cost reasons. ZipLoom still runs the same checks before every deploy.
Walkthrough
- Open Settings → AWS and find Connect your AWS account.
- Enter an AWS Access Key ID and Secret Access Key for a root or admin-level user, choose the Region, then choose Connect AWS account.
- ZipLoom uses those credentials once to create a scoped deploy sub-user in your account, then discards them. Deploys run as that sub-user, and the card shows its name and region.
- The card shows when the connection was Last verified. Choose Re-verify at any time to check it still works; a problem is shown on the card.
- Pick AWS as a project's deploy target and deploy as usual.
Options & Settings
- Disconnect stops ZipLoom deploying to the account. You are asked to confirm first.
- Use the show/hide control to check the secret key before you submit it.
Edge Cases & Gotchas
- Business plan and above. On lower plans the page shows the plan you need.
- The admin-level credentials are not kept. If you later rotate or delete that AWS user, existing deploys are unaffected because they use the scoped deploy user.
- Self-healing rollback covers AWS deploys, because there is an earlier version to restore.
- Everything ZipLoom creates lives in your AWS account and is billed by AWS to you.
Related
Deploy Targets · Self-Host Export · Self-Healing Deploys
Last updated: 2026-11-10 · docs v5.1
Account
MCP Servers
What it is. A registry of external MCP (Model Context Protocol) servers for your organization (/settings/mcp-servers): a name, an address and an optional access token for each one.
Why it matters. It keeps the connection details for the tools your team uses in one place, with tokens encrypted, so they are ready when ZipLoom's AI features can call them.
Where it stands today. Registered servers are stored and managed here. ZipLoom's AI features do not call them yet. This page is not how a coding agent connects to ZipLoom; for that, see MCP For Coding Agents.
Walkthrough
- Open Settings → MCP servers.
- Add a server: Name (for example Notion), URL (a full address such as
https://mcp.example.com/mcp), an optional Auth token and an optional description.
- Choose Test connection. A green tick means something answered at that address; a red cross means nothing did.
- The server appears under Registered servers. A Token Saved badge shows when a token is stored.
Options & Settings
- Edit or delete a server at any time.
- Adding, editing, testing and deleting need an Owner or Admin. Every member can see the list.
Edge Cases & Gotchas
- Test connection checks reachability only. It confirms the address responds; it does not complete an MCP handshake or prove the token works.
- Tokens are encrypted at rest and never shown again after you save them.
- A URL that is not a full address is refused: "the URL must be a full address such as https://mcp.example.com/mcp."
- "No MCP servers registered yet." until you add one.
Related
MCP For Coding Agents · AI Builder · Settings & Workspace
Last updated: 2026-11-10 · docs v5.1
Account
Founding Member Plans
What it is. Founding-member pricing and one-time offers (/founding).
Walkthrough
- Open Founding. 2. Choose an option and go through founding checkout. Options include First Thread — Founding Solo, Founding Guild, packs, and White-Glove Weave (a hands-on migration by a person: repositories moved, environment configured, DNS pointed). 3. Review What is included before checkout.
Options & Settings
- Founding (subscription-style) and one-time checkouts (billing.foundingCheckout / oneTimeCheckout).
Edge Cases & Gotchas
- Founding offers are limited-time/limited-seat by nature, and founding prices stay below public pricing.
- The Keep-Your-Apps Guarantee: ZipLoom does not hold your applications hostage to your subscription.
- Payments are handled by Stripe; ZipLoom never sees your card.
Related
Billing & Plans
Last updated: 2026-11-09 · docs v5.0
Account
Two-Factor Authentication
What it is. Every ZipLoom account must use two-factor authentication (2FA): a code from an authenticator app at each sign-in, with one-use recovery codes as a backup.
Why it matters. ZipLoom holds the keys to your hosting, repositories and databases. A stolen password alone should never be enough to reach them.
Walkthrough
- After you first sign in, ZipLoom asks you to set up 2FA.
- Scan the QR code with an authenticator app (Google Authenticator, Authy, 1Password, Bitwarden or Microsoft Authenticator), or enter the secret key by hand.
- Enter the 6-digit code to confirm.
- Save your recovery codes somewhere safe, such as a password manager. They are shown only once. Choose I've saved them.
Options & Settings
- Add another authenticator at any time from Settings.
- Generate new recovery codes if you have none left.
Edge Cases & Gotchas
- 2FA is enforced on the server, not only in the browser.
- Your only authenticator cannot be removed. To replace it, add the new one first, then remove the old one.
- Each recovery code works once. Without your authenticator or a recovery code you cannot sign in.
Related
Settings & Workspace · Team & Roles
Last updated: 2026-11-09 · docs v5.0
Account
MCP For Coding Agents
What it is. Connect your coding agent to ZipLoom over MCP. Claude Code, Codex, Cursor, VS Code, Windsurf and Gemini CLI can list your projects, run scans, read findings and fixes, and check deploy status and live posture without leaving the editor.
Why it matters. Your agent writes the code. Connected to ZipLoom, it can also check it and fix the right thing first.
Walkthrough
- In Settings → API Keys, create a key and choose IDE Agent (MCP).
- Put the key in your environment so config files never contain it.
- Follow the setup for your agent at
ziploom.dev/mcp: the server URL plus an Authorization: Bearer header.
- Try: "Scan this project with ZipLoom and fix whatever the attack chains say to fix first."
Options & Settings
- Agents can also read GitHub through ZipLoom with read-only tools. Changes they propose become requests that an Owner or Admin approves before anything runs.
- Each fix comes with a paste-ready brief, the same as Copy Fix For Your AI Tool.
Edge Cases & Gotchas
- A key only reaches its own organization; a project in another organization reads as not found.
- Agent-started scans reuse a scan already in progress and are limited to five an hour per organization, so a looping agent cannot use up the scans your CI depends on.
- Revoke a key in Settings and it stops working at once.
- This is different from MCP Servers, which registers outside servers for ZipLoom's own AI builder.
Related
CI, API Keys & The PR Check (Secure) · MCP Servers · Attack Chains (Secure)
Last updated: 2026-11-09 · docs v5.0
Free Tools & Research
Free Tools & Quick Scan
What it is. Free checks at ziploom.dev/tools that anyone can run on a public website with no account: security headers, email security, exposed source maps, CORS, and page checks for SEO, AI search readiness, accessibility and legal pages. Quick Scan runs the four security checks in one go.
Why it matters. A two-minute check catches the launch-day mistakes that are easy to miss, before a customer or attacker does.
Walkthrough
- Go to
ziploom.dev/tools and pick a tool, or open Quick Scan.
- Enter your site's address and confirm the human check.
- Read the findings, each with how to fix it.
- Share a report link if you want someone else to see the results.
Options & Settings
- Security checks: Security Headers, Email Security (SPF, DKIM, DMARC), Source-Map Exposure, CORS Checker.
- Page checks: SEO Checker, AI Search Readiness, Accessibility Checker, Legal Pages Checker.
Edge Cases & Gotchas
- The checks are passive: they read what any visitor can see and send no attack traffic.
- The free tools show findings and fixes, not a grade. For a grade you can publish, use Launch Readiness.
- A short human check (Cloudflare Turnstile) protects the tools from automated abuse.
Related
Launch Readiness Grade · Live Posture Scan (Operate)
Last updated: 2026-11-09 · docs v5.0
Free Tools & Research
Launch Readiness Grade
What it is. A letter grade for a website you own, at ziploom.dev/tools/launch-readiness. Prove the domain is yours, get the grade with copy-paste fixes, and publish a shareable result with a Verified badge.
Why it matters. A grade you can share tells customers and partners your launch was checked, and the fixes show exactly what to change to raise it.
Walkthrough
- Open Launch Readiness and enter your domain.
- Prove you own it with a DNS record, a file on the site, or a meta tag.
- Get your grade and the fix snippets.
- Publish the result to get a share link and a Verified badge.
Options & Settings
- Shared grades are re-checked daily, so a published badge keeps proving itself.
- The page checks (SEO, accessibility and so on) are not part of this grade, by design.
Edge Cases & Gotchas
- You can only grade a domain you can prove you control.
- If the site changes, the daily re-check updates the published grade.
Related
Free Tools & Quick Scan · Trust Badge & Share Security Proof (Own)
Last updated: 2026-11-09 · docs v5.0
Free Tools & Research
Research: AI Coding Benchmark & Security Index
What it is. Two public research pages from ZipLoom. The Weekly AI-Coding Benchmark (ziploom.dev/research/ai-coding-benchmark) asks several AI models the same plain web-app tasks each week and scans the code they write. The Monthly AI Code Security Index (ziploom.dev/research/ai-code-security-index) summarizes what the month showed across public repositories, ZipLoom's own data, and public vulnerability sources.
Why it matters. It shows, with evidence, how often AI-written code ships with security problems, and which kinds.
Walkthrough
- Open the benchmark to see, per model, how many answers were scanned and the share with no critical or high finding.
- Open the Index for the latest monthly edition. Sign in, or enter your email, to read the full report.
Options & Settings
- Any figure based on fewer than 10 results is shown as Withheld rather than as a rate.
- Index editions are reviewed and published by a person, and cannot be changed once published.
Edge Cases & Gotchas
- Public repositories are grouped by which AI tool they are set up for. That shows the tool configured, not that the code was written by AI.
- No repository, owner or code is ever named or shown. Only totals are stored.
- The repositories are what GitHub search returns, not a random sample.
Related
Free Tools & Quick Scan · Security Scanning (Secure)
Last updated: 2026-11-09 · docs v5.0