Developers · MCP
Let your AI client run the scanner, not guess the answer.
The Model Context Protocol is how tools talk to AI clients like Claude, Cursor, and custom agents. ProofKit’s endpoint speaks the core of it - initialize, tools/list, tools/call - over plain HTTP. Three tools work with no token at all: scan a resume against the 23-check rubric, search live job listings, list templates. Three more act on your saved resumes and need a bearer token you mint below while signed in.
Endpoint: https://proofkit.store/mcp (POST JSON-RPC 2.0).
The tools
Six calls. Three are public.
Schemas are served live by tools/list - this table is the human version.
| Tool | Token? | What it does |
|---|---|---|
proofkit_scan_resume | No | Runs the deterministic 23-check ATS scan: score, 5 category scores, and every failed check with its fix. No AI involved. |
proofkit_search_jobs | No | Live remote-friendly listings from a public board: title, company, tags, salary, apply URL. Cached 5 minutes. |
proofkit_list_templates | No | The 15 template ids, which are free and which need Pro. |
proofkit_list_resumes | Yes | Your saved resumes: id, title, template, last updated. |
proofkit_read_resume | Yes | One resume by id, including its full data. |
proofkit_write_resume | Yes | Creates a saved resume on your account. Free plan keeps its 1-resume cap - the tool tells you instead of silently failing. |
Your token
Mint a bearer token, paste it in your client.
Sign in first (email link, no password). The token is tied to your account and expires after 90 days.
Client config (bearer header form):
{
"mcpServers": {
"proofkit": {
"url": "https://proofkit.store/mcp",
"headers": { "Authorization": "Bearer PASTE_YOUR_TOKEN" }
}
}
}
If your client only accepts a URL with no custom headers, it cannot attach the bearer yet - the three public tools still work without one.
Questions
What people ask about the endpoint.
What works without signing in?
Will it change my resume in the builder?
proofkit_write_resume creates a saved copy in your account list; it never overwrites a resume, and it does not touch the draft stored in your browser. Edit it in the builder afterwards if you want.Why does the scan beat asking the model directly?
Scan first. Automate second.
The endpoint is for workflows. The builder is where the resume actually gets better.