quickstart.
fizzgig is a catalogue of audit tools your AI editor can call directly through MCP. One install command, one paste, and your editor can scan for secret leaks, RLS misconfigs, dependency CVEs, and more — without you ever leaving the chat.
Fizzgig audits the code, not the builder — so it works on AI-built apps too. Sync your Lovable / Bolt project to GitHub (or export it), open the repo in Cursor or Claude Code, and run the audit there. It's the fastest way to catch the classics those builders ship by default: Supabase RLS left open, the service-role key leaked to the browser, routes with no auth. full guide →
Run this in your terminal. It detects your editor, asks for a Fizzgig key, writes the right config file, and tells you how to reload — everything below this card is the manual fallback.
$ npx @fizzgig-ai/installnode 18+ required · source on npm
three steps, ~60 seconds.
get an API key
▸ no account yet? request beta access →
run the installer
$ npx @fizzgig-ai/installThe installer detects which editor you have, asks which to install Fizzgig into (multi-select if you have several), prompts for the API key from step 1, then writes the right config file per editor. Your existing MCP servers are preserved — Fizzgig is added alongside, not blasted over.reload your editor + try a tool
“What Fizzgig tools do you have access to?”
It should list your tools. (If it says it has none, the server isn't connected — re-check the Authorization header and that you fully reloaded.) Then run one on something real:
“Use fizzgig__secret_leak_finder on the contents of lib/ and workers/.”
The AI calls Fizzgig, gets back ranked findings, surfaces them with copy-paste fix suggestions. Every run also writes an audit-log row you can view at /dashboard/runs.
what you have access to.
Six tools land in your editor on install. Eighteen more roll out as the catalogue opens (paid tier — see /pricing).
when: Before every commit, before publishing any repo, after adding a new SDK.
when: Before any production migration, when adding a new table, when rewriting a policy.
when: After every npm install, before every release, on every CI PR.
when: After every npm install. Database adds advisories daily — even a static lockfile can become vulnerable overnight.
when: After every new fetch() / axios call to a new domain.
when: After other tools run (the map gets richer with each emit). Before any deploy.
common issues.
The installer says “no supported MCP-aware editors detected”
My editor doesn't see Fizzgig after the install
“Invalid or missing execution secret” / 401
The tool returned 56KB and tripped my AI client's output cap
verbose: true if you need the full envelope back (debugging, dashboard inspection). Otherwise responses sit comfortably under 10KB even for multi-finding scans.next.
something not working? file an issue at github.com/RascalRebelCreative/fizzgig/issues — every coverage gap or edge case is one we want to close before public launch.