Jobs Every Hour MCP server
# Jobs Every Hour MCP server
An [MCP](https://modelcontextprotocol.io) server that lets Claude, ChatGPT, Cursor and other AI assistants search open jobs from
[Jobs Every Hour](https://jobseveryhour.com/): jobs read every hour from employers' own career pages (Greenhouse, Lever, Ashby,
Workday, iCIMS and other hiring systems). Closed jobs are removed, and every job links to a page whose Apply button goes to the
employer's own application.
It reads the site's [free JSON API](https://jobseveryhour.com/api/), which needs no key and no account.
## Tools
| Tool | What it returns |
| --- | --- |
| `latest_jobs` | The newest open jobs (optionally remote only) |
| `jobs_by_skill` | Jobs that ask for a skill: Python, React, SQL, Kubernetes, Salesforce... |
| `jobs_by_category` | Jobs in a field: engineering, healthcare, sales, operations... |
| `jobs_by_location` | Jobs in a country, region or city: united-states, europe, london... |
| `jobs_by_company` | Jobs at one employer, from its own career page |
| `list_feeds` | The skills, categories, locations and companies you can ask for, with job counts |
Each tool returns up to 30 of the newest jobs with title, employer, place, remote flag, advertised pay, posting date and link.
## Install
Needs Node.js 18 or newer.
**Claude Desktop** (`claude_desktop_config.json`), **Cursor** (`.cursor/mcp.json`) and other clients that start a local server:
```json
{
"mcpServers": {
"jobseveryhour": {
"command": "npx",
"args": ["-y", "github:Simba0315/jobseveryhour-mcp"]
}
}
}
```
**Claude Code:**
```bash
claude mcp add jobseveryhour -- npx -y github:Simba0315/jobseveryhour-mcp
```
Then ask, for example: "Find remote Python jobs posted this week", "What is Stripe hiring for?" or "Show nursing jobs in Texas".
## Notes
- Feeds change every hour; the server caches each one for an hour.
- Use of the data: free for personal and commercial projects with a link back to the job page or to jobseveryhour.com.
- Issues and ideas: open an issue in this repository.
## License
MIT
TDQS
Scored across 6 tools
Each tool targets a distinct facet of the job corpus (recency, skill, category, location, company) plus a metadata tool, so an agent can map intent to tool without guessing. The only near-overlap is latest_jobs versus a filtered query, but its description makes the recency purpose unambiguous.
Five of six tools follow a clear jobs_by_<dimension> pattern with a final list_feeds verb_noun tool. latest_jobs deviates slightly but remains readable and conventional in snake_case, so the set is predictable overall.
Six tools is well within the ideal range and each maps to a genuine search dimension; none feels redundant or bolted on. list_feeds also earns its place by exposing valid filter values.
The surface covers listing and faceted filtering but has no way to retrieve a single job's full detail by id, nor keyword/title or salary/date-range search, which are common job-search needs. Agents can still do broad discovery, but deeper lookups or refinement are dead ends.