Use the Workerbee MCP server

Connect AI agents to Workerbee through the MCP server to search candidate rankings, manage jobs, and query talent data from Claude, Cursor, or Claude Code.

The Workerbee MCP server enables AI agents to interact directly with Workerbee’s talent intelligence functionality via the standardized Model Context Protocol (MCP). Agents can search Success Profiles, manage jobs and candidate pipelines, rank and compare candidates, and pull evidence-backed match details for any role.

Workerbee also provides a dedicated MCP server scoped to each job or Success Profile — see Use a job-scoped MCP server. For plugins and skills built on top of this server, see Agentic IDEs and CLIs.

Tools

Full tool surface, grouped for readability. Tool availability depends on your plan/tier — a Builder-tier key may not see every tool a Standard/Enterprise key does:

Group	Tools
Discovery / read	`whoami`, `get_portfolio_summary`, `get_job_context`, `list_my_jobs`, `get_success_profile`, `get_success_profile_history`
Candidates & matching	`list_candidates`, `find_candidate_rank`, `match_candidates`, `get_matched_profile_details`, `get_worker_passport`, `read_worker_skills`, `read_worker_experience`
Act on a role	`create_job`, `propose_success_profile`, `apply_success_profile_proposal`, `set_evaluation_weights`, `set_location_preference`, `invite_candidate`, `move_candidate`, `remove_candidate`
Audit & governance	`get_decision_audit`
Workforce data	`create_internal_worker`, `update_worker_passport`, `bulk_update_worker_passports`, `suggest_worker_skill_updates`, `suggest_worker_experience_updates`, `delete_worker_skills`
Support	`create_support_ticket`

bulk_update_worker_passports and delete_worker_skills affect multiple records at once, or are irreversible — use with care.

Before you begin

You need:

A Workerbee account (invite-only — see Intelligence Console quickstart).
An API key, generated from the Intelligence Console — Settings → API keys. Copy it once; the Console can rotate or revoke it later.
An MCP-compatible client (Claude Code, Claude Desktop, Cursor, or any other client that supports remote HTTP MCP servers).

Every config below uses the same production endpoint, https://mcp.workerbee.ai/api/v1/mcp, and the same bearer-token pattern (Authorization: Bearer YOUR_API_KEY) — a different auth model from LLM Connectors, which use an OAuth sign-in flow instead of a pasted API key. That page is the hiring-flow product surface inside a consumer AI assistant; this page is for a developer wiring an agent/IDE directly to the API with a key they hold.

For the easiest setup, install the Workerbee Claude Code plugin instead — it bundles the MCP server, skills, and slash commands. The manual configuration below is for users who only want the MCP server.

Add the MCP server. Run:

bash

claude mcp add-json workerbee-mcp \
'{"type": "http",
"url": "https://mcp.workerbee.ai/api/v1/mcp",
"headers": {"Authorization": "Bearer YOUR_API_KEY"}}'

Replace YOUR_API_KEY with your WORKERBEE_API_KEY.

Check the status. Restart Claude Code, then run /mcp:

> /mcp
⎿ MCP Server Status

• workerbee-mcp: ✓ connected

Test the server with a task prompt, e.g.:

Rank the candidates for the Senior Data Engineer job against its Success Profile and summarize the top 5 with evidence.

Add the MCP server. Settings → Developer → Edit Config:

json

{
  "mcpServers": {
    "workerbee": {
      "type": "http",
      "url": "https://mcp.workerbee.ai/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Replace YOUR_API_KEY with your WORKERBEE_API_KEY.

Check the status. Restart Claude Desktop — a hammer (MCP) icon should appear on the new chat screen with Workerbee’s tools listed.

Test the server with the same task prompt as above.

Rank the candidates for the Senior Data Engineer job against its Success Profile and summarize the top 5 with evidence.

For the easiest setup, install the Workerbee Cursor plugin instead — it bundles the MCP server, skills, and slash commands. The manual configuration below is for users who only want the MCP server.

Add the MCP server. Create .cursor/mcp.json in your project root:

json

{
  "mcpServers": {
    "workerbee": {
      "type": "http",
      "url": "https://mcp.workerbee.ai/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Replace YOUR_API_KEY with your WORKERBEE_API_KEY.

Check the status. Cursor Settings → MCP — you should see the server and its tool list.

Add Workerbee rules (optional, but keeps the server well-behaved). Create .cursor/rules/workerbee.mdc:

### Tool Usage for Talent Queries
- Always use the `workerbee` MCP and call `get_skill_instructions` before any other tool.
- Do not assemble a workflow from the raw tool list — follow the playbook a matched skill returns.

### Error Handling
- If a tool returns FEATURE_NOT_IN_PLAN or DEMO_EXPIRED, relay the message verbatim and stop.
- Never fabricate or simulate a ranking, match, or candidate result.

### Data Handling
- Candidate identities stay anonymized (initials or bee-styled handles) until consent is confirmed by the server.

Test the server — Cmd/Ctrl+I to open the Agent chat, then the same task prompt as above.

Any MCP-compatible client that supports remote HTTP servers can connect using the same endpoint and bearer token:

json

{
  "mcpServers": {
    "workerbee": {
      "type": "http",
      "url": "https://mcp.workerbee.ai/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Replace YOUR_API_KEY with your WORKERBEE_API_KEY, then confirm the connection through that client’s own MCP status view before testing with a prompt that queries jobs, candidates, or rankings.

Was this page helpful?