Open Greenhouse MCP

What benmonopoli/open-greenhouse-mcp MCP server does

The benmonopoli/open-greenhouse-mcp MCP server connects an MCP-compatible client to Greenhouse recruiting systems. It combines direct API coverage with higher-level recruiting workflows, so an agent can answer pipeline questions or perform multi-step tasks without requiring each API request to be managed separately.

The server covers candidate, application, job, offer, interview, and related Greenhouse data through Harvest API tools. It also includes job board operations for public listings and application submission. Optional webhook and ingestion capabilities are documented for event-driven workflows and partner integrations.

Profile selection controls the available tools and write access:

• read-only provides 103 tools and disables writes.
• recruiter provides 127 tools and allows recruiter-safe writes.
• full provides 181 tools and enables all listed tools.

The recruiter profile includes actions such as rejecting, advancing, hiring, moving, tagging, adding notes, managing attachments and interviews, handling prospects, and running bulk operations. It does not include administrative areas such as job creation, user management, custom-field configuration, candidate deletion, or webhook management.

Tools and capabilities

The benmonopoli/open-greenhouse-mcp MCP server includes screening and sourcing functions for assembling a candidate profile, extracting resume text, searching resumes with required, preferred, and excluded keywords, and finding candidates by structured fields. It can also retrieve applications added since a specified date and group them by job.

Composite workflow tools combine multiple API calls into recruiting-oriented results, including:

• Pipeline summaries grouped by stage, with candidate names and days in stage
• Stale applications and interviews that lack scorecards
• Conversion, hiring, rejection, and time-in-stage metrics
• Source effectiveness and time-to-hire statistics
• Candidate lookup by name or exact email
• Resume and attachment retrieval
• Bulk rejection, tagging, and advancement with rate-limit handling

Destructive actions require explicit IDs rather than inferred targets. Write activity can include an audit attribution through GREENHOUSE_ON_BEHALF_OF, while access remains constrained by the permissions of the configured Greenhouse API credential.

Setup and configuration

Install the Python package with pip install open-greenhouse-mcp, then configure the executable open-greenhouse-mcp as an MCP stdio server in the client. The README gives configurations for Claude Desktop and Cursor. Python 3.10 or newer is supported.

Set at least one of these connection variables:

• GREENHOUSE_API_KEY: a Greenhouse Harvest API key
• GREENHOUSE_BOARD_TOKEN: the URL slug for a Greenhouse job board

GREENHOUSE_TOOL_PROFILE selects full, recruiter, or read-only; full is the default. GREENHOUSE_READ_ONLY=true is also described as a read-only shorthand. For write audit attribution, set GREENHOUSE_ON_BEHALF_OF to a Greenhouse user ID.

Logging can be adjusted with GREENHOUSE_LOG_LEVEL, using debug, info, warning, or error, and output can be redirected with GREENHOUSE_LOG_FILE. Structured JSON logs include API method, status, latency, and related event data.

Limitations and notes

The benmonopoli/open-greenhouse-mcp MCP server depends on Greenhouse credentials and the permissions attached to them. A read-only profile is recommended for initial connectivity and behavior checks. The number of available tools and whether writes are possible depend on the selected profile.

The documented transport is stdio. Supported clients listed in the project documentation include Claude Desktop, Claude Code, and Cursor. Bulk operations are rate-limited to remain within API limits, and destructive operations do not choose their targets automatically. The project is released under the MIT License.