Grey Ridge Jobs MCP
# Greenhouse Jobs MCP Server
[](https://github.com/rezearcher/jobs-mcp-server/actions/workflows/smoke.yml)
[](https://modelcontextprotocol.io)
[](package.json)
A **Model Context Protocol** server that gives any MCP client — Claude Desktop, Cursor, Cline,
Continue, or your own agent — live, normalized job postings straight from a company's
**Greenhouse** board (the most-requested ATS feed), plus seven more applicant-tracking systems
in one clean schema.
One tool call, Greenhouse + seven more ATS platforms, one clean schema:
> **Greenhouse · Lever · Ashby · SmartRecruiters · Workday · Recruitee · Workable · BambooHR**
Give it a bare company name (`stripe`), a domain (`stripe.com`), or an explicit board
(`greenhouse:stripe`, `workday:nvidia.wd5:NVIDIAExternalCareerSite`) — the server detects the ATS
and board token for you and returns postings with stable field names across every platform.
- **Zero dependencies.** No MCP SDK, no transitive package tree — the server is one file plus a
vendored adapter core. It installs and boots instantly under `npx`, and drags nothing into your
agent's trust boundary. (For an agent tool, the supply chain *is* the attack surface.)
- **Zero keys, zero account.** Every source is a public ATS JSON endpoint.
- **Normalized schema.** `title · company · location · remote · employment_type · department ·
posted_at (ISO) · salary · url · source_platform` — same shape whoever the employer uses.
- **Auto-detect.** Pass a slug or domain; the ATS and token are resolved for you.
- **Plus Hacker News "Who is hiring?"** The `search_hn_hiring` tool reads the current month's thread
into the same schema, and finds next month's thread by itself when it is posted.
## Install
Requires Node.js ≥ 18.
### Claude Desktop
Add to `claude_desktop_config.json` (Settings → Developer → Edit Config):
```json
{
"mcpServers": {
"jobs": {
"command": "npx",
"args": ["-y", "github:rezearcher/jobs-mcp-server"]
}
}
}
```
### Cursor / Cline / Continue
Point the client at the same command (`npx -y github:rezearcher/jobs-mcp-server`) as a stdio
server. In Cursor: Settings → MCP → Add new server → type `command`.
### From source
```bash
git clone https://github.com/rezearcher/jobs-mcp-server
cd jobs-mcp-server
node server.mjs # speaks MCP over stdio
npm test # real client handshake + live-data roundtrip
```
## Tools
| Tool | What it does |
|------|--------------|
| `search_jobs` | Live normalized postings for one or more companies. Args: `companies` (string[]), `maxPerCompany` (default 25, max 200), `includeDescription` (default false). |
| `detect_ats` | Resolve a company slug/domain to `{platform, token}` without pulling the full list. |
| `list_platforms` | The 8 supported platforms and how to address each. No network call. |
| `search_hn_hiring` | Live postings from the **current** month's Hacker News *"Ask HN: Who is hiring?"* thread, parsed into the same schema. Args: `query` (keyword filter), `remoteOnly` (default false), `maxResults` (default 50, max 200), `includeDescription`. The monthly thread id is auto-discovered — it rotates on the 1st of every month, so you never re-point it. |
| `get_premium_access` | Describe the metered depth endpoints beyond the free reads above — a normalized cross-ATS **compensation** layer and a per-company **hiring-intent signal** — served pay-per-call over x402 (USDC on Base, no account, the caller's wallet pays the 402 inline). No network call, no payment. |
### Free vs. metered
The four search tools above are **free** and read the live boards directly. Two derived layers are
heavier to keep correct — pay-transparency text drifts, cross-ATS comp needs normalizing — so they
are served **pay-per-call** over the [x402](https://x402.org) rail (USDC on Base, no API key): a few
tenths of a cent per call. Call `get_premium_access` for the exact endpoints, prices, and payment
handshake, or hit the free preview first: `https://assetfactory-jobs-x402.sigrunner.workers.dev/jobs/preview?companies=stripe.com`.
### Example
Ask your assistant: *"Use the jobs tool to list open engineering roles at Stripe and Palantir."*
The model calls:
```json
{ "name": "search_jobs",
"arguments": { "companies": ["greenhouse:stripe", "lever:palantir"], "maxPerCompany": 25 } }
```
and gets back a `jobs[]` array plus a `resolutions[]` log showing how each company was resolved.
Or, to sweep the whole Hacker News hiring thread: *"Find remote Rust roles in this month's HN
who-is-hiring thread."* →
```json
{ "name": "search_hn_hiring",
"arguments": { "query": "rust", "remoteOnly": true } }
```
## Addressing a company
| Form | Example | Notes |
|------|---------|-------|
| bare slug | `stripe` | auto-detected across platforms |
| domain | `stripe.com` | auto-detected |
| explicit | `greenhouse:stripe`, `lever:palantir`, `ashby:openai` | skips detection |
| Workday | `workday:nvidia.wd5:NVIDIAExternalCareerSite` or a `*.myworkdayjobs.com` URL | explicit only |
## Why a maintained server (and not a one-off script)
ATS JSON shapes drift and companies migrate between vendors. This server is the public face of a
continuously **maintained** adapter layer: each platform is isolated so a shape change touches one
adapter, and a scheduled canary (`.github/workflows/smoke.yml`) exercises a live board on every push
and daily, failing loudly the moment an adapter stops returning rows. That upkeep is the point — it
is why the feed keeps working after the day you install it.
## License
MIT © Grey Ridge Signals Group LLC. Job content belongs to the respective employers and their ATS
providers; this tool only retrieves and normalizes publicly listed postings. Respect each source's
terms and rate limits.
## Related tools (same maintained engine)
Prefer a plain CLI, or a company-level signal instead of an MCP tool?
- **[ats-jobs-scraper-cli](https://github.com/rezearcher/ats-jobs-scraper-cli)** — the same 8-ATS feed as a zero-dependency command-line tool.
- **[unified-jobs-scraper-cli](https://github.com/rezearcher/unified-jobs-scraper-cli)** — adds LinkedIn public postings and cross-source dedup.
- **[company-enrichment-cli](https://github.com/rezearcher/company-enrichment-cli)** — domain → one clean company record (firmographics, contacts, tech stack) with a rolled-up hiring-intensity signal from the same ATS engine; sales/prospecting buying-intent rather than a raw job feed.
TDQS
Scored across 5 tools
Each tool has a distinct role: detect_ats is a cheap probe, search_jobs fetches company postings, search_hn_hiring pulls a different source, list_platforms is static metadata, and get_premium_access is informational. The only mild overlap is that search_jobs already auto-detects the ATS, which makes detect_ats somewhat redundant unless the caller explicitly wants the probe — but the descriptions call this out clearly.
All five names follow a clean verb_noun snake_case convention (detect_ats, search_jobs, list_platforms, search_hn_hiring, get_premium_access) with no mixed styles or vague verbs.
Five tools is well within the sweet spot and each one earns its place: ATS detection, ATS job search, platform enumeration, HN thread search, and premium-tier discovery. Nothing feels padded or missing at the count level.
Core discovery workflows are covered — search across 7 ATS platforms plus HN, with a unified schema that includes url, salary, and remote flags. Minor gaps: search_jobs doesn't advertise keyword/location/remote filters (unlike search_hn_hiring), and there's no job-detail or company-listing operation, though agents can work around these via the returned urls.