Serveur MCP Greenhouse pour les workflows de recrutement

[project]
name = "greenhouse-recruiting-mcp"
version = "0.1.0"
description = "Read-mostly MCP server exposing Greenhouse Harvest API to Claude / MCP-compatible clients."
readme = "README.md"
requires-python = ">=3.11"
license = { text = "MIT" }
authors = [
    { name = "ooligo authors" }
]
dependencies = [
    "mcp>=0.9.0",
    "httpx>=0.27.0",
    "pydantic>=2.7.0",
]

[project.optional-dependencies]
dev = [
    "pytest>=8.0.0",
    "pytest-asyncio>=0.23.0",
    "ruff>=0.5.0",
]

[project.scripts]
greenhouse-recruiting-mcp = "greenhouse_recruiting_mcp.server:main"

[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"

[tool.hatch.build.targets.wheel]
packages = ["src/greenhouse_recruiting_mcp"]

[tool.ruff]
line-length = 100
target-version = "py311"

# Greenhouse recruiting MCP server

A read-mostly MCP server exposing the Greenhouse Harvest API as tools to Claude Desktop, Claude Code, or any MCP-compatible client. Six read tools cover daily recruiter questions; one cautious write tool (`note_stage_stuck`) adds an internal note.

**This is a scaffold, not a runtime-tested production server.** The tool implementations are written against the Harvest API's documented shape, but the recruiting engineer is responsible for verifying each tool against a Greenhouse staging tenant before flipping production credentials. The disclaimer is repeated in `server.py`'s module docstring.

## Install

```bash
cd apps/web/public/artifacts/mcp-server-greenhouse-recruiting/
pip install -e .
# or
uv pip install -e .
```

The package exposes a `greenhouse-recruiting-mcp` CLI entrypoint.

## Environment variables

### `GREENHOUSE_API_KEY` (required)

The Harvest API key from Greenhouse → Configure → Dev Center → API Credential Management.

When generating the key, pick the **minimum scope** needed:

- `Get` permission on `Candidates`, `Applications`, `Jobs`, `Users`, `Departments`.
- `Post` permission on `Candidate Notes` (only — required for `note_stage_stuck`).

Wider scopes silently turn the server into a higher-blast-radius surface. If you find yourself adding a write permission for a new tool later, document the per-tool justification in `server.py`'s `TOOL_REGISTRY` docstring.

### `GREENHOUSE_USER_ID_FOR_ON_BEHALF_OF` (required)

The Greenhouse user ID that writes will be attributed to. Find this in the Greenhouse user URL when you're logged in as the user (e.g. `app.greenhouse.io/users/123456` → user ID is `123456`).

This is required even if you only use the read tools — the server validates it at startup so writes can't slip through unattributed later.

## MCP client registration

### Claude Desktop

Add to `claude_desktop_config.json` (location varies by OS — `~/Library/Application Support/Claude/` on macOS, `%APPDATA%\Claude\` on Windows):

```json
{
  "mcpServers": {
    "greenhouse-recruiting": {
      "command": "uv",
      "args": ["run", "greenhouse-recruiting-mcp"],
      "cwd": "/absolute/path/to/mcp-server-greenhouse-recruiting",
      "env": {
        "GREENHOUSE_API_KEY": "...",
        "GREENHOUSE_USER_ID_FOR_ON_BEHALF_OF": "..."
      }
    }
  }
}
```

Restart Claude Desktop. The seven tools should appear in the tools panel.

### Claude Code

In your project's `.claude/settings.local.json`:

```json
{
  "mcpServers": {
    "greenhouse-recruiting": {
      "command": "uv",
      "args": ["run", "greenhouse-recruiting-mcp"],
      "cwd": "/absolute/path/to/mcp-server-greenhouse-recruiting",
      "env": {
        "GREENHOUSE_API_KEY": "...",
        "GREENHOUSE_USER_ID_FOR_ON_BEHALF_OF": "..."
      }
    }
  }
}
```

### Other MCP clients (Cursor, Continue, etc.)

Most accept the same `command + args + env` shape. Refer to the client's MCP documentation.

## Sanity-check invocation

Before the first real use, run the server against a Greenhouse staging tenant (paid Greenhouse customers can request a staging environment from their CSM).

```bash
GREENHOUSE_API_KEY=staging_key \
GREENHOUSE_USER_ID_FOR_ON_BEHALF_OF=staging_user_id \
greenhouse-recruiting-mcp --help
```

(The CLI is the MCP stdio server; `--help` is not implemented in this scaffold. The intent is to confirm the package installed and the entrypoint resolved.)

For per-tool validation, the recommended flow is:

1. Register the server with Claude Desktop pointed at staging credentials.
2. Ask Claude to call each tool with known staging inputs and verify the responses match what you see in the staging Greenhouse UI.
3. Only after every tool is verified, swap to production credentials.

## Security model

- **Auth.** Greenhouse API key as Basic-auth username, empty password. Greenhouse's documented pattern.
- **Writes.** Only `note_stage_stuck` mutates state. Attributed via `On-Behalf-Of` header so the Greenhouse audit log shows the recruiting engineer's user, not just the API key.
- **Rate limit.** Token-bucket at 40 req/10s by default (Greenhouse ceiling is 50 req/10s). Lower if other systems share the API key.
- **PII in MCP responses.** The server returns the data the API returns — including candidate names and emails. The calling Claude session is downstream; the session's data-handling posture is the recruiter's responsibility. Don't paste session transcripts into shared Slack channels; don't log raw responses to your own audit table.
- **Audit log.** The server logs every tool call to stderr at INFO level with PII-stripped arguments. Recruiting engineer is responsible for capturing stderr into a durable audit log (e.g. via systemd journal, Docker log driver, or a wrapping script that tees to a file).

## Known limits — numbered TODO before production use

The scaffold is honest about what it doesn't do yet. Treat each as a numbered TODO to close before broad production use.

1. **Not runtime-tested.** Every tool needs validation against a Greenhouse staging tenant. The smoke check in this README is a credentials check, not a per-tool validation.
2. **Pagination max page count.** The async iterator caps at 50 pages per call (5,000 records at 100/page). For tenants with very large candidate volumes, the cap needs raising or replacement with a streaming pattern.
3. **No request retry on transient errors.** The 429 handler retries once with a 2-second backoff. Other 5xx errors propagate; the recruiting engineer wraps the calls if more retry resilience is needed.
4. **Stage-name matching is exact-string.** "Phone Screen" and "Phone screen" are different. The tool surfaces this clearly enough but does not normalize.
5. **No multi-tenant support.** One server instance, one Greenhouse account. Multi-tenant requires non-trivial reshape (per-call credential injection, tenant-aware audit logging).
6. **Activity feed parsing is partial.** `get_candidate_history` returns the activity items the Harvest endpoint exposes; some interaction types (system actions, automated emails) may be undertyped. Expand the schema as the team finds gaps.
7. **No tests.** A pytest suite for the rate limiter and the Link-header parser is the obvious first addition; full integration tests against a staging tenant are the second.

## What this server intentionally does NOT do

- **No `delete_*` tools.** Deletes happen in the Greenhouse UI, with the audit trail that produces.
- **No candidate-state mutations** (move stages, advance to offer, reject). Those are recruiter decisions and need explicit per-tool justification — adding them would compromise the read-mostly posture.
- **No bulk send / outbound email.** Outreach belongs in a sourcing tool with proper unsubscribe handling, not in an MCP read-tool surface.
- **No PII normalization** (no name-redaction, no email-hashing in responses). The server returns what Greenhouse returns; downstream redaction is the recruiter's responsibility.

## Adding a new tool

If you need a new tool:

1. Add a Pydantic input schema and an async implementation in `server.py`.
2. Register it in `TOOL_REGISTRY` with a clear description.
3. If it's a write tool, document the per-tool justification in the function's docstring (see `note_stage_stuck` for the template). Confirm the `On-Behalf-Of` attribution flows through.
4. Validate against staging.
5. Update this README's tool list.

The scaffold's structure makes this a 30-60 minute change per tool. The discipline is in the per-tool justification step — it's the only thing that prevents the read-mostly posture from drifting.

"""Greenhouse recruiting MCP server."""

__version__ = "0.1.0"

"""
Greenhouse recruiting MCP server.

Exposes seven tools to MCP-compatible clients (Claude Desktop, Claude Code,
Cursor, etc.) backed by the Greenhouse Harvest API. Six are read; one is
the cautious write (`note_stage_stuck`).

NOT runtime-tested against a live Greenhouse tenant. The tool dispatch
implementations are written against Greenhouse's documented Harvest API
shape (https://developers.greenhouse.io/harvest.html as of 2026-Q2), but
the production deployment requires the recruiting engineer to verify each
tool against a staging tenant before flipping the production credentials.

Security model:
  - Auth: Greenhouse API key (Harvest scope) via Basic auth, key as username,
    empty password — Greenhouse's documented pattern.
  - Writes: only `note_stage_stuck` mutates state; uses `On-Behalf-Of`
    header for audit attribution.
  - Rate limit: token-bucket (default 40 req/10s; Greenhouse ceiling is
    50 req/10s per API key per IP).
  - Pagination: cursor-based on most endpoints; the implementations loop
    until no `Link: rel="next"` header is present.
  - Audit: every tool call logged to stderr at INFO level with tool name,
    parameters (PII-stripped), and response status. Recruiting engineer
    is responsible for capturing these into a durable audit log.
"""

from __future__ import annotations

import asyncio
import logging
import os
import time
from collections.abc import AsyncIterator
from typing import Any

import httpx
from mcp.server import Server
from mcp.types import Tool, TextContent
from pydantic import BaseModel, Field

logger = logging.getLogger(__name__)


# --- Configuration ------------------------------------------------------------

GREENHOUSE_API_BASE = "https://harvest.greenhouse.io/v1"
DEFAULT_RATE_LIMIT_PER_10S = 40  # Greenhouse documented ceiling: 50.
DEFAULT_TIMEOUT_S = 30.0


def _require_env(name: str) -> str:
    val = os.environ.get(name)
    if not val:
        raise RuntimeError(
            f"Required env var {name} not set. The Greenhouse MCP server cannot start "
            f"without credentials. See README.md for setup."
        )
    return val


# --- Rate limiter -------------------------------------------------------------


class TokenBucket:
    """Simple token-bucket rate limiter, async-safe."""

    def __init__(self, rate: int, per_seconds: float) -> None:
        self.rate = rate
        self.per_seconds = per_seconds
        self.tokens = float(rate)
        self.last_refill = time.monotonic()
        self._lock = asyncio.Lock()

    async def acquire(self) -> None:
        async with self._lock:
            now = time.monotonic()
            elapsed = now - self.last_refill
            self.tokens = min(self.rate, self.tokens + elapsed * (self.rate / self.per_seconds))
            self.last_refill = now
            if self.tokens >= 1:
                self.tokens -= 1
                return
            wait = (1 - self.tokens) * (self.per_seconds / self.rate)
        await asyncio.sleep(wait)
        await self.acquire()


# --- Greenhouse client --------------------------------------------------------


class GreenhouseClient:
    """Thin async wrapper around the Harvest API."""

    def __init__(
        self,
        api_key: str,
        on_behalf_of_user_id: str,
        rate_limit_per_10s: int = DEFAULT_RATE_LIMIT_PER_10S,
    ) -> None:
        self.api_key = api_key
        self.on_behalf_of_user_id = on_behalf_of_user_id
        self.bucket = TokenBucket(rate=rate_limit_per_10s, per_seconds=10.0)
        self._client = httpx.AsyncClient(
            base_url=GREENHOUSE_API_BASE,
            auth=(api_key, ""),
            timeout=DEFAULT_TIMEOUT_S,
            headers={"User-Agent": "greenhouse-recruiting-mcp/0.1.0"},
        )

    async def close(self) -> None:
        await self._client.aclose()

    async def _request(
        self,
        method: str,
        path: str,
        *,
        params: dict[str, Any] | None = None,
        json: dict[str, Any] | None = None,
        attribute_write: bool = False,
    ) -> httpx.Response:
        await self.bucket.acquire()
        headers: dict[str, str] = {}
        if attribute_write:
            headers["On-Behalf-Of"] = self.on_behalf_of_user_id
        resp = await self._client.request(method, path, params=params, json=json, headers=headers)
        if resp.status_code == 429:
            # Greenhouse does not return Retry-After on 429. Back off conservatively.
            await asyncio.sleep(2.0)
            return await self._request(
                method, path, params=params, json=json, attribute_write=attribute_write
            )
        resp.raise_for_status()
        return resp

    async def paginate(
        self,
        path: str,
        params: dict[str, Any] | None = None,
        *,
        max_pages: int = 50,
    ) -> AsyncIterator[dict[str, Any]]:
        """Yield each item from a paginated Harvest endpoint."""
        url: str | None = path
        page = 0
        while url is not None and page < max_pages:
            resp = await self._request("GET", url, params=params if page == 0 else None)
            for item in resp.json():
                yield item
            link = resp.headers.get("Link", "")
            url = _next_url_from_link_header(link)
            page += 1


def _next_url_from_link_header(link_header: str) -> str | None:
    """Parse RFC-5988 Link header for rel=next."""
    if not link_header:
        return None
    for part in link_header.split(","):
        section = part.split(";")
        if len(section) < 2:
            continue
        url = section[0].strip().strip("<>")
        rel = section[1].strip()
        if rel == 'rel="next"':
            # Greenhouse returns absolute URL; httpx follows but we want the path.
            if url.startswith(GREENHOUSE_API_BASE):
                return url[len(GREENHOUSE_API_BASE):]
            return url
    return None


# --- Pydantic schemas ---------------------------------------------------------


class ListCandidatesInStageInput(BaseModel):
    job_id: int = Field(..., description="Greenhouse job ID")
    stage_name: str = Field(..., description="Stage name as it appears in Greenhouse")
    stale_after_days: int | None = Field(
        None, description="Optional filter: candidates last touched more than N days ago"
    )


class GetCandidateHistoryInput(BaseModel):
    candidate_id: int = Field(..., description="Greenhouse candidate ID")


class ListJobsOpenInput(BaseModel):
    department: str | None = Field(None, description="Optional department name filter")


class GetFunnelForJobInput(BaseModel):
    job_id: int = Field(..., description="Greenhouse job ID")


class ListJobsStalledInput(BaseModel):
    stale_after_days: int = Field(
        7, description="A job is stalled if no candidate progressed in this many days"
    )


class SearchCandidatesByAttributeInput(BaseModel):
    custom_field_name: str
    value: str


class NoteStageStuckInput(BaseModel):
    candidate_id: int
    note_body: str = Field(..., description="The note text. Visible only internally.")


# --- Tool implementations -----------------------------------------------------


async def list_candidates_in_stage(
    client: GreenhouseClient, args: ListCandidatesInStageInput
) -> list[dict[str, Any]]:
    """Return candidates currently in a named stage on a given job."""
    out: list[dict[str, Any]] = []
    cutoff = (
        time.time() - (args.stale_after_days or 0) * 86400 if args.stale_after_days else None
    )
    async for app in client.paginate(
        f"/jobs/{args.job_id}/applications", params={"per_page": 100, "status": "active"}
    ):
        current_stage = app.get("current_stage", {})
        if (current_stage.get("name") or "").strip() == args.stage_name.strip():
            last_activity = app.get("last_activity_at")
            if cutoff and last_activity:
                if time.mktime(time.strptime(last_activity[:19], "%Y-%m-%dT%H:%M:%S")) > cutoff:
                    continue
            out.append(
                {
                    "candidate_id": app.get("candidate_id"),
                    "application_id": app.get("id"),
                    "applied_at": app.get("applied_at"),
                    "last_activity_at": last_activity,
                    "current_stage": current_stage.get("name"),
                }
            )
    return out


async def get_candidate_history(
    client: GreenhouseClient, args: GetCandidateHistoryInput
) -> dict[str, Any]:
    """Return a candidate's stage history."""
    candidate_resp = await client._request("GET", f"/candidates/{args.candidate_id}")
    candidate = candidate_resp.json()
    activity_resp = await client._request(
        "GET", f"/candidates/{args.candidate_id}/activity_feed"
    )
    activity = activity_resp.json()
    return {
        "candidate_id": candidate.get("id"),
        "name": candidate.get("first_name", "") + " " + candidate.get("last_name", ""),
        "current_application_ids": [app.get("id") for app in candidate.get("applications", [])],
        "activities": [
            {
                "type": a.get("subject"),
                "at": a.get("created_at"),
                "by": a.get("user"),
            }
            for a in (activity.get("activities") or [])
        ],
    }


async def list_jobs_open(
    client: GreenhouseClient, args: ListJobsOpenInput
) -> list[dict[str, Any]]:
    """List all open jobs."""
    out: list[dict[str, Any]] = []
    async for job in client.paginate(
        "/jobs",
        params={"status": "open", "per_page": 100},
    ):
        if args.department and (job.get("departments") or [{}])[0].get("name") != args.department:
            continue
        out.append(
            {
                "job_id": job.get("id"),
                "name": job.get("name"),
                "department": (job.get("departments") or [{}])[0].get("name"),
                "opened_at": job.get("opened_at"),
                "closed_at": job.get("closed_at"),
                "hiring_managers": [
                    h.get("name") for h in (job.get("hiring_team", {}).get("hiring_managers") or [])
                ],
            }
        )
    return out


async def get_funnel_for_job(
    client: GreenhouseClient, args: GetFunnelForJobInput
) -> dict[str, int]:
    """Return candidate count per stage for a job."""
    counts: dict[str, int] = {}
    async for app in client.paginate(
        f"/jobs/{args.job_id}/applications", params={"per_page": 100, "status": "active"}
    ):
        stage = (app.get("current_stage", {}).get("name") or "unknown").strip()
        counts[stage] = counts.get(stage, 0) + 1
    return counts


async def list_jobs_stalled(
    client: GreenhouseClient, args: ListJobsStalledInput
) -> list[dict[str, Any]]:
    """List jobs where no candidate has progressed in N days."""
    cutoff = time.time() - args.stale_after_days * 86400
    stalled: list[dict[str, Any]] = []
    async for job in client.paginate("/jobs", params={"status": "open", "per_page": 100}):
        latest_activity = 0.0
        async for app in client.paginate(
            f"/jobs/{job['id']}/applications", params={"per_page": 100, "status": "active"}
        ):
            la = app.get("last_activity_at")
            if la:
                try:
                    t = time.mktime(time.strptime(la[:19], "%Y-%m-%dT%H:%M:%S"))
                    if t > latest_activity:
                        latest_activity = t
                except ValueError:
                    continue
        if latest_activity > 0 and latest_activity < cutoff:
            stalled.append(
                {
                    "job_id": job.get("id"),
                    "name": job.get("name"),
                    "days_since_progress": int((time.time() - latest_activity) / 86400),
                }
            )
    return stalled


async def search_candidates_by_attribute(
    client: GreenhouseClient, args: SearchCandidatesByAttributeInput
) -> list[dict[str, Any]]:
    """Search candidates by a custom field value."""
    out: list[dict[str, Any]] = []
    async for c in client.paginate("/candidates", params={"per_page": 100}):
        for cf in c.get("custom_fields", []) or []:
            if cf.get("name") == args.custom_field_name and str(cf.get("value")) == args.value:
                out.append(
                    {
                        "candidate_id": c.get("id"),
                        "name": c.get("first_name", "") + " " + c.get("last_name", ""),
                        "matched_field": cf.get("name"),
                        "matched_value": cf.get("value"),
                    }
                )
                break
    return out


async def note_stage_stuck(
    client: GreenhouseClient, args: NoteStageStuckInput
) -> dict[str, Any]:
    """
    Add an internal note to a candidate. The single write tool exposed.

    Per-tool justification:
      - Required to log "Claude flagged this candidate as stage-stuck" so the
        action is visible in the audit trail and not silent.
      - No candidate-state mutation (does not move stages, does not send
        emails, does not change scorecards).
      - Attributed via On-Behalf-Of header so the Greenhouse audit log
        shows the recruiting-engineer user, not just the API key.
    """
    body = {
        "user_id": int(client.on_behalf_of_user_id),
        "body": args.note_body,
        "visibility": "private",  # internal note, not visible to candidate
    }
    resp = await client._request(
        "POST",
        f"/candidates/{args.candidate_id}/activity_feed/notes",
        json=body,
        attribute_write=True,
    )
    return {"status": "ok", "note": resp.json()}


# --- MCP server wiring --------------------------------------------------------

TOOL_REGISTRY: dict[str, tuple[type[BaseModel], Any, str]] = {
    "list_candidates_in_stage": (
        ListCandidatesInStageInput,
        list_candidates_in_stage,
        "List candidates currently in a named stage on a given job. Optionally filter by staleness.",
    ),
    "get_candidate_history": (
        GetCandidateHistoryInput,
        get_candidate_history,
        "Return a candidate's stage history and activity feed.",
    ),
    "list_jobs_open": (
        ListJobsOpenInput,
        list_jobs_open,
        "List open jobs. Optional department filter.",
    ),
    "get_funnel_for_job": (
        GetFunnelForJobInput,
        get_funnel_for_job,
        "Return candidate counts per stage for a single job.",
    ),
    "list_jobs_stalled": (
        ListJobsStalledInput,
        list_jobs_stalled,
        "List jobs where no candidate has progressed in N days.",
    ),
    "search_candidates_by_attribute": (
        SearchCandidatesByAttributeInput,
        search_candidates_by_attribute,
        "Search candidates by custom field name and value.",
    ),
    "note_stage_stuck": (
        NoteStageStuckInput,
        note_stage_stuck,
        "Write tool: add a private internal note to a candidate. Audit-attributed via On-Behalf-Of.",
    ),
}


def build_server() -> Server:
    server = Server("greenhouse-recruiting-mcp")

    api_key = _require_env("GREENHOUSE_API_KEY")
    on_behalf_of = _require_env("GREENHOUSE_USER_ID_FOR_ON_BEHALF_OF")
    client = GreenhouseClient(api_key=api_key, on_behalf_of_user_id=on_behalf_of)

    @server.list_tools()
    async def _list_tools() -> list[Tool]:
        return [
            Tool(
                name=name,
                description=desc,
                inputSchema=schema.model_json_schema(),
            )
            for name, (schema, _, desc) in TOOL_REGISTRY.items()
        ]

    @server.call_tool()
    async def _call_tool(name: str, arguments: dict[str, Any]) -> list[TextContent]:
        if name not in TOOL_REGISTRY:
            return [TextContent(type="text", text=f"Unknown tool: {name}")]
        schema, fn, _ = TOOL_REGISTRY[name]
        try:
            args = schema.model_validate(arguments)
        except Exception as exc:
            logger.warning("Tool %s called with invalid args: %s", name, exc)
            return [TextContent(type="text", text=f"Invalid arguments: {exc}")]

        # Audit: log tool call with PII-light args (drop free-text body for note tool).
        audit_args = arguments.copy()
        if name == "note_stage_stuck":
            audit_args["note_body"] = f"<{len(arguments.get('note_body', ''))} chars>"
        logger.info("Tool call: %s args=%s", name, audit_args)

        try:
            result = await fn(client, args)
        except httpx.HTTPStatusError as exc:
            logger.warning("Tool %s HTTP error: %s", name, exc)
            return [
                TextContent(
                    type="text",
                    text=f"Greenhouse API error {exc.response.status_code}: {exc.response.text[:500]}",
                )
            ]
        except Exception as exc:
            logger.exception("Tool %s failed", name)
            return [TextContent(type="text", text=f"Tool failed: {exc}")]

        # Result returned as JSON-shaped text content; the calling Claude session parses it.
        import json
        return [TextContent(type="text", text=json.dumps(result, default=str, indent=2))]

    return server


def main() -> None:
    """Entry point for `greenhouse-recruiting-mcp` CLI."""
    logging.basicConfig(
        level=logging.INFO,
        format="%(asctime)s %(levelname)s %(name)s %(message)s",
    )
    from mcp.server.stdio import stdio_server

    async def _run() -> None:
        server = build_server()
        async with stdio_server() as (read_stream, write_stream):
            await server.run(
                read_stream,
                write_stream,
                server.create_initialization_options(),
            )

    asyncio.run(_run())


if __name__ == "__main__":
    main()