Greenhouse MCP
Production-ready MCP server for Greenhouse ATS with 175 tools for recruiting teams — manage candidates, applications, jobs, interviews, and hiring pipelines. Role-based profiles (full/recruiter/read-only), composite workflow tools for pipeline views, analytics, candidate search, and bulk operations.
README
open-greenhouse-mcp
<!-- mcp-name: io.github.benmonopoli/greenhouse-mcp -->
Production-ready MCP server for Greenhouse, designed for recruiters and hiring teams.
Most Greenhouse MCP servers mirror the API endpoint by endpoint. This one is built for recruiting teams: safe defaults, role-based profiles, and workflow tools that turn multi-step API operations into single actions.
Choose a Profile
| Profile | Tools | Can write? | Recommended for |
|---|---|---|---|
read-only |
97 | No | First-time setup, reporting, hiring managers |
recruiter |
121 | Yes (safe ops) | Day-to-day recruiting work |
full |
175 | Yes (all) | Admins, ops, advanced automation |
Quick Start
pip install open-greenhouse-mcp
Add to your MCP client config (Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json, Cursor: Settings > MCP):
{
"mcpServers": {
"greenhouse": {
"command": "open-greenhouse-mcp",
"env": {
"GREENHOUSE_API_KEY": "your-harvest-api-key",
"GREENHOUSE_TOOL_PROFILE": "read-only"
}
}
}
}
Start in read-only mode to validate connectivity and tool behaviour, then switch to recruiter or full when you need write access.
Your API key is in Greenhouse under Configure > Dev Center > API Credential Management.
What You Can Ask
- "Show me the pipeline for our Senior Engineer role"
- "Who needs my attention this week?"
- "What are our conversion rates for the Backend Intern role?"
- "Find Sarah Chen and pull up her resume"
- "Which sources are actually producing hires?"
- "Bulk reject everything inactive for 30+ days on the Account Manager role"
See more examples with full output.
See it in action

Safety
- Access is limited by your Greenhouse API key permissions
- Read-only profile is recommended for first setup
- Destructive actions require explicit IDs — the server never infers targets
- Write operations support audit attribution via
GREENHOUSE_ON_BEHALF_OF - Bulk actions are rate-limited to stay within API limits
Compatibility
| Client | Status |
|---|---|
| Claude Desktop | Supported |
| Claude Code | Supported |
| Cursor | Supported |
| Transport | stdio |
| Python | 3.10+ |
Startup
When the server starts, it logs its configuration:
open-greenhouse-mcp v0.3.0
Profile: recruiter | Tools: 121 | Writes: recruiter-safe | APIs: harvest, ingestion
What's Included
- Recruiter workflow tools — 13 composite tools for pipeline views, analytics, search, and bulk operations
- Harvest API coverage — 148 tools across candidates, applications, jobs, offers, interviews, and more
- Job Board API — 13 tools for public job listings and application submission
- Optional webhooks and ingestion — 14 tools for event-driven workflows and partner integrations
Reference
Composite Tools
High-level tools that combine multiple API calls into single operations.
| Tool | What it does |
|---|---|
pipeline_summary |
Full pipeline view — candidates grouped by stage with names and days-in-stage |
candidates_needing_action |
Find stale applications and interviews missing scorecards |
stale_applications |
Applications with no activity for N days, sorted by stalest |
pipeline_metrics |
Conversion rates, hire/rejection rates, time-in-stage per stage |
source_effectiveness |
Which candidate sources produce the best hire rates |
time_to_hire |
Average, median, min, max days from application to hire |
bulk_reject |
Reject multiple applications in one call with rate-limit handling |
bulk_tag |
Tag multiple candidates in one call |
bulk_advance |
Advance multiple applications to next stage |
search_candidates_by_name |
Find candidates by first or last name |
search_candidates_by_email |
Look up a candidate by exact email |
read_candidate_resume |
Download and return a candidate's most recent resume |
download_attachment |
Download any Greenhouse attachment by URL |
Profile Details
Recruiter includes all read tools, all composite workflows, and recruiter-safe writes: reject, advance, hire, move, tag, notes, attachments, interviews, prospects, and bulk operations. It excludes job creation, user management, custom field configuration, candidate deletion, and webhook management.
Read-only skips all write operations. GREENHOUSE_READ_ONLY=true also works as a shorthand.
Configuration
| Variable | Required | Description |
|---|---|---|
GREENHOUSE_API_KEY |
Yes* | Harvest API key |
GREENHOUSE_BOARD_TOKEN |
Yes* | Job board URL slug. *At least one required |
GREENHOUSE_TOOL_PROFILE |
No | full (default), recruiter, or read-only |
GREENHOUSE_ON_BEHALF_OF |
No | Greenhouse user ID for write audit trail |
GREENHOUSE_LOG_LEVEL |
No | debug, info, warning (default), error |
GREENHOUSE_LOG_FILE |
No | Log file path (defaults to stderr) |
Logging
Structured JSON logging for observability. Set GREENHOUSE_LOG_LEVEL=info to enable:
{"ts": "2026-04-14T12:31:58", "level": "info", "event": "api_call", "method": "GET", "url": "...", "status": 200, "latency_ms": 245.0}
More Documentation
- API Reference — Full tool breakdown by category
- Usage Examples — Real conversations with full output
- Advanced Setup — Webhook receiver, ingestion API, board-token mode
- Development — Contributing, testing, project structure
Feedback
- Bugs and features: Open an issue
- Questions: Start a discussion
- Security: See SECURITY.md
- Contributing: See CONTRIBUTING.md
License
MIT License -- Ben Monopoli. See LICENSE.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。