Recruiter Roles MCP server

The Model Context Protocol is an open standard that lets AI assistants call external tools and data sources. Our MCP server exposes the Recruiter Roles job board as a set of tools, so an assistant can search jobs, look up companies, and pull market stats on your behalf — no glue code required.

It is a thin, read-only layer over our v1 REST API. Same data, same API key, same rate limit — just spoken instead of curled.

The server speaks Streamable HTTP (the current MCP transport) at a single endpoint:

https://recruiterroles.com/api/mcp

Stdio-only clients (older setups) can bridge to it with the mcp-remote npm package — see the configs below.

Authenticate with the same rr_live_ key you use for the REST API. Issue one instantly on the API access page. Two ways to present it:

Seven read-only tools, one per data type:

Claude.ai (web) & Claude Desktop — custom connector: open Settings → Connectors → Add custom connector, name it “Recruiter Roles”, and paste your URL with the key appended:

https://recruiterroles.com/api/mcp?key=rr_live_YOUR_KEY

Claude Desktop — config file (claude_desktop_config.json) keeps the key out of the URL by using a header:

{
  "mcpServers": {
    "recruiter-roles": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://recruiterroles.com/api/mcp",
        "--header", "Authorization: Bearer rr_live_YOUR_KEY"
      ]
    }
  }
}

Restart Claude, then ask: “Use Recruiter Roles to find remote technology recruiter jobs paying over $120k.”

Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json (per-project):

{
  "mcpServers": {
    "recruiter-roles": {
      "url": "https://recruiterroles.com/api/mcp",
      "headers": {
        "Authorization": "Bearer rr_live_YOUR_KEY"
      }
    }
  }
}

Cursor picks it up automatically — check Settings → MCP for a green dot, then use it from the Agent.

ChatGPT (connectors): in a Developer-mode / custom connector field, add the server URL with your key appended:

https://recruiterroles.com/api/mcp?key=rr_live_YOUR_KEY

OpenAI Responses API: pass it as a remote MCP tool:

curl https://api.openai.com/v1/responses \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5",
    "tools": [{
      "type": "mcp",
      "server_label": "recruiter_roles",
      "server_url": "https://recruiterroles.com/api/mcp",
      "headers": { "Authorization": "Bearer rr_live_YOUR_KEY" },
      "require_approval": "never"
    }],
    "input": "Find remote technology recruiter jobs paying over 120k."
  }'

VS Code (Copilot) — .vscode/mcp.json:

{
  "servers": {
    "recruiter-roles": {
      "type": "http",
      "url": "https://recruiterroles.com/api/mcp",
      "headers": { "Authorization": "Bearer rr_live_YOUR_KEY" }
    }
  }
}

Any Streamable HTTP client: point it at https://recruiterroles.com/api/mcp with an Authorization: Bearer header. Stdio-only clients: wrap it with mcp-remote as shown in the Claude Desktop example.

Quick test from your terminal

# List the available tools
curl -s "https://recruiterroles.com/api/mcp" \
  -H "Authorization: Bearer rr_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Or run the official inspector: npx @modelcontextprotocol/inspector, then connect to the URL above.

MCP tool calls share your REST quota: 1,000 requests per day per key, resetting at midnight UTC. Each tool call counts as one request. Exceeding the limit returns a clear rate-limit error the assistant will relay.

The same policy as the REST API applies: any page or product that publishes data obtained through this server must include a visible, crawlable link back to Recruiter Roles. Keys may be suspended if the backlink is removed.