cvjson-mcp
v1.1.1
Published
MCP server for cv.json — search people who are open to work, and fetch & validate machine-readable CVs, from any AI assistant.
Maintainers
Readme
cvjson-mcp
An MCP server for cv.json — give any AI assistant the power to find people who are open to work and read their structured CVs.
cv.json is the open, machine-readable CV format from
cvjson.com / FreeCV. Every FreeCV
portfolio publishes one at livelink.cv/<name>/cv.json.
Tools
| Tool | What it does |
|------|--------------|
| search_open_to_work | Search a live index of real people open to work by skills, role/title, location and recruiter facets — work type (remote/hybrid/onsite), employment type, spoken languages, visa sponsorship, minimum years of experience — with sorting and paging. Returns candidate cards + each person's cv.json URL. |
| fetch_cv | Fetch a person's full machine-readable CV by slug or URL. |
| validate_cv | Validate any document against the official cv.json schema. |
| get_cvjson_schema | Return the cv.json JSON Schema. |
Only already-public data is returned; contact details never are.
Install
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"cvjson": {
"command": "npx",
"args": ["-y", "cvjson-mcp"]
}
}
}Restart Claude Desktop — the cvjson tools appear in the 🔌 menu.
Any MCP client (Cursor, Windsurf, custom)
Run it as a stdio server:
npx -y cvjson-mcpHosted HTTP endpoint (no install)
Clients that speak the Streamable HTTP transport can connect directly:
https://freecv.org/api/mcpTry it
Once connected, ask your assistant things like:
"Find me React developers who are open to work." "Fetch the cv.json for
ashleyand summarise their experience." "Is this cv.json valid?" (paste one)
How it works
fetch_cv,validate_cvandget_cvjson_schemarun locally —validate_cvchecks against the always-current canonical schema atcvjson.com.search_open_to_workis served by the hosted cv.json index (it needs the live database of who's open to work). Because the hosted server executes that tool, it also owns its contract: this package asks it for the tool's own definition at startup, so new search facets appear without a package upgrade. The bundled definition is the offline fallback.
Self-host or point at your own deployment:
CVJSON_MCP_ENDPOINT=https://your-host/api/mcp \
CVJSON_SCHEMA_URL=https://your-host/schema/v1.json \
npx -y cvjson-mcpRequirements
Node.js ≥ 18.
