Give Claude eyes, ears & hands
The full 120-module scanner inside your AI assistant — plus 24 tools that let it see the rendered page, hear what's breaking in production, and prove each fix worked. 100% free on your own machine — Claude Code, Cursor, Windsurf, any MCP agent.
Free · every tool · your machine, your keys
claude mcp add gatetest -- npx -y @gatetest/mcp-server
Can't run npm — or want GateTest inside claude.ai on web and mobile? The hosted endpoint runs the scans on our infrastructure and keeps your scan history.
Cancel anytime · API key emailed instantly · The local server stays free forever
Install anywhere — 30 seconds, any environment
No terminal? No npm? No problem. The hosted endpoint reaches every Claude user.
claude.ai web & mobile zero install
Settings → Connectors → Add custom connector
URL: https://gatetest.ai/api/mcp
Claude Desktop App one-click
Download the extension, double-click it — done
⬇ Download gatetest.mcpbOr paste into Settings → Developer → Edit Config: { "mcpServers": { "gatetest": { "url": "https://gatetest.ai/api/mcp" } } }
Claude Code CLI full 24 tools
Local install — unlocks scan_local, run_tests, query_db, stream_logs
claude mcp add gatetest -- npx -y @gatetest/mcp-server
Cursor / Windsurf / Cline / Zed
MCP settings → Add server (URL or command, both work)
https://gatetest.ai/api/mcp
Hosted endpoint: free tools work with no key; add Authorization: Bearer gtmcp_... to unlock premium. Filesystem tools (scan_local, run_tests, query_db, stream_logs, http_request) need the local install.
24 tools — every one free on the local server
The table shows the local server (npx @gatetest/mcp-server). On the hosted endpoint, five tools work with no key (check_health, list_modules, get_badge, scan_url, scan_repo); hosted scans, AI fix/diagnose, and scan history need the $29/mo key.
| Tool | What it does | Access |
|---|---|---|
| check_health | Verify GateTest engine is operational | FREE |
| list_modules | List all 120 modules with descriptions | FREE |
| get_badge | Get embeddable README badge for any repo | FREE |
| scan_url | Quick scan any live URL via hosted API | FREE |
| scan_local (quick) | 41-module quick scan — syntax, lint, secrets, codeQuality, and more | FREE |
| scan_local (standard/full/smart) | 45-module standard, 88-module full, or diff-aware smart scan | FREE |
| run_module | Run one specific module against a path | FREE |
| fix_issue | AI-driven auto-fix for a specific finding | FREE |
| explain_finding | Forensic-tier Claude diagnosis per finding | FREE |
| compose_pr | Render a PR body for a set of fixes | FREE |
| capture_screenshot | 👁 Eyes — screenshot any live URL or localhost so the AI sees the rendered page | FREE |
| get_visual_diff | 👁 Eyes — baseline vs current visual diff to catch UI regressions | FREE |
| run_live_checks | 👂 Ears — runtime errors, console warnings, API health on any URL | FREE |
| get_production_errors | 👂 Ears — top Sentry / Datadog / Rollbar errors with file:line attribution | FREE |
| verify_fix | 🤝 Hands — re-scan changed files for hard pass/fail proof the fix worked | FREE |
| run_tests | 🤝 Hands — auto-detect + run the project's test suite (Jest/Vitest/pytest/cargo/go) | FREE |
| stream_logs | 🤝 Hands — tail a running process or log file in real time (up to 60s) | FREE |
| query_db | 🤝 Hands — read-only SQL/NoSQL queries (Postgres/MySQL/SQLite/MongoDB/Redis) | FREE |
| http_request | 🤝 Hands — call any API with auth headers, follow redirects, inspect responses | FREE |
| audit_log | Query past local scans in the memory store | FREE |
| compare_repos | Cross-repo prior-art lookup via memory store | FREE |
| get_report | Retrieve full result of the last scan this session | FREE |
| scan_repo | Scan any public git repo via the hosted API | FREE |
| resolve_stack_trace | Minified/bundled stack trace → original file:line via source maps | FREE |
| blame_regression | Which git commit introduced this line — read-only, ranks candidates across a stack trace | FREE |
See the rendered page
capture_screenshot — see what the rendered page actually looks like. Works on localhost, staging, and production.
get_visual_diff — baseline vs current comparison so Claude spots regressions before you do.
Hear what's breaking
get_production_errors — pull your top Sentry, Datadog, or Rollbar errors with file:line attribution so Claude fixes what prod says is broken, first.
run_live_checks — runtime error sweep, console warnings, and API health against any URL.
Prove the fix worked
verify_fix — re-run the relevant modules on changed files. Pass/fail verdict so Claude knows the fix actually worked.
run_tests — auto-detect and run Jest, Vitest, pytest, cargo test, or go test. Structured pass/fail per test.
stream_logs — tail a running process or log file live for up to 60s while Claude is debugging.
query_db — read-only SQL and NoSQL queries (Postgres, MySQL, SQLite, MongoDB, Redis) without leaving the session.
http_request — call any API with auth headers, inspect responses, follow redirects. Closes the loop: scan → fix → test → verify → done.
FAQ
What's free?
The entire local server. Every tool — full-suite scans, screenshots, production errors, run_tests, fix_issue — runs 100% free on your own machine via npx @gatetest/mcp-server (AI tools use your own Anthropic key). The $29/mo key unlocks the HOSTED endpoint: use GateTest from claude.ai web/mobile or locked-down machines where you can't run npm, plus hosted scan history. On the hosted endpoint, check_health, list_modules, get_badge, scan_url, and scan_repo work with no key at all.
How do I get my API key?
Subscribe below. Your key (format: gtmcp_xxx) is emailed to you within seconds of checkout completing.
How do I add it to Claude Code?
Free local server (every tool): claude mcp add gatetest -- npx -y @gatetest/mcp-server Hosted endpoint (claude.ai web/mobile): add https://gatetest.ai/api/mcp as a custom connector, with Authorization: Bearer gtmcp_xxx for the paid tools.
Does the key expire?
Your key is valid as long as your subscription is active. Cancel anytime — the key stops working at the end of your billing period.
Can I use it with Cursor, Windsurf, or other MCP clients?
Yes. GateTest MCP follows the MCP spec — any client that supports stdio transport works. Set GATETEST_API_KEY in the environment for that server.
Which AI model runs my fixes — and who pays for it?
You choose, and you pay Anthropic directly (bring-your-own-key). AI fixes run on YOUR ANTHROPIC_API_KEY — calls go straight from your machine to Anthropic, never through our servers, and you control the spend. Pick the model per call: sonnet (Claude Sonnet 5, default — fast and cheapest), opus (Opus 4.8 — deeper reasoning), or fable (Fable 5 — the most capable model Anthropic ships, ~3.3x Sonnet cost). No other QA tool lets you do this.
Ready?
Free on your machine: npx -y @gatetest/mcp-server. Want it in claude.ai web/mobile? Subscribe and your key arrives in seconds.
Cancel anytime · API key emailed instantly · Local server free forever