MCP server · REST API · RSS · llms.txt
Startup jobs, for your AI agent
Connect Claude, ChatGPT, Cursor or an agent of your own to startup.jobs. Search 274,000 live jobs at 71,800 startups, see who's hiring and benchmark salaries, all inside the conversation.
Find remote full-time jobs at startups posted this week.
YipitData · India
Smartsheet · Bengaluru, India
Gympass · Portugal
These are the newest matches. Want me to filter by role or country, or pull the full description of one?
Works with the agents you already use
- Claude Web, desktop, mobile
- ChatGPT Custom connector
- Claude Code One command
- Cursor One-click install
- Codex CLI and IDE
- VS Code One-click install
- Windsurf Cascade
- Gemini CLI settings.json
Plus any other MCP client, over Streamable HTTP. Setup for each.
One dataset, three ways in
Whether you are job hunting with an assistant, building a product, or giving your own agent a new tool.
Job seekers
Ask Claude or ChatGPT for startup jobs the way you would ask a friend. Your assistant searches live listings, reads full descriptions and compares salaries, then sends you to the job page to apply.
Connect your assistantDevelopers
Build job widgets, newsletters, bots and research tools on a free REST API with an OpenAPI spec, cursor pagination and the same filters the MCP tools use. RSS feeds if you want zero code.
Read the API docsAgent builders
Give a personal or product agent a startup-jobs tool in one line of config. Eight read-only tools with typed schemas, structured results and attribution baked in, so your agent cites the source.
Browse the toolsWays to connect
Four are live today. Three more are on the roadmap, and what you tell us you would use decides the order.
MCP server
For assistants and agents. Streamable HTTP, no account needed, eight read-only tools.
https://api.startup.jobs/mcp
REST API
For code. JSON endpoints for jobs, companies, roles and locations, with an OpenAPI spec and free keys.
https://api.startup.jobs/v1/jobs
RSS feeds
For readers and automation tools. The 50 newest jobs, filterable by role, company or remote. No key.
https://startup.jobs/feeds/jobs
llms.txt
For models reading the web. What startup.jobs is, who it is for, and where everything lives, in plain text.
https://startup.jobs/llms.txt
On the roadmap
Post jobs via MCP
PlannedLet startups publish and manage listings from their own agent, payment included.
CLI
PlannedA command-line client for scripts and shell-first agents, with the same tools as MCP.
Webhooks
PlannedA ping when a new job matches your filters, so your agent reacts instead of polling.
Would use one of these? Tell us what you're building.
What you can ask
Plain language in, structured data out. Each question maps to one of the eight tools.
-
“Find remote senior Rails jobs at startups posted this week.”
search_jobs -
“What is Stripe hiring for right now?”
get_company_jobs -
“Show me the full description for that second job.”
get_job -
“What do product designers earn at US startups?”
salary_benchmarks -
“How has hiring for AI engineers changed over the last year?”
job_trends -
“Which countries have the most open data roles?”
list_countries
Built to be used, not scraped
We want every agent to be able to work with startup jobs. A few rules keep that sustainable.
- Read-only
- Agents search and read. Applying stays with people, on the job page.
- Link back
- Every result carries its startup.jobs URL. Show it when you show the job.
- Recent by default
- Free access covers the last 14 days, which is what job seekers need. The archive is for full-access partners.
- Fair limits
- 20 requests a minute without a key. Building something bigger? Talk to us.
The details are in the API terms.
Questions, answered
Something missing? Email us.
How do I find startup jobs with ChatGPT or Claude?
Connect the Startup Jobs MCP server to your assistant (Settings → Connectors in Claude, Apps & Connectors in ChatGPT) and ask it in plain language: "find remote senior Rails jobs at startups posted this week". The assistant calls our search_jobs tool, gets live listings back, and links you to each job on startup.jobs. Setup takes about a minute and needs no account.
What is an MCP server, and why does a job board have one?
The Model Context Protocol (MCP) is the open standard AI assistants use to call external tools. Our MCP server gives any MCP client eight read-only tools over our listings: search jobs, fetch a full description, look up a company and its openings, list roles and countries, and run hiring-trend and salary-benchmark queries. It means your agent works with structured, current startup job data instead of guessing from the open web.
Can my agent apply to jobs for me?
No. Every tool is read-only by design: agents can search and read, and each job comes with its startup.jobs URL so you apply there yourself. We do this on purpose. Applications written and sent by agents are bad for candidates and for the startups hiring them.
Do I need an API key?
Not for the MCP server: anonymous access works with all eight tools for listings from the last 14 days. The REST API needs a free key from any startup.jobs account. A key on the MCP server attributes your usage and, when your account has full access, unlocks the complete archive, structured salary data and longer trend series.
Which agents and tools are supported?
Anything that speaks MCP over Streamable HTTP. We document setup for Claude, ChatGPT, Claude Code, Cursor, Codex, VS Code, Windsurf, and Gemini CLI, and the generic config works for other editors, personal agents and your own code. Agents that cannot use MCP can call the REST API or read llms.txt.
Is it free?
Yes. The MCP server and the REST API are free for personal use and for building things, with a 14-day listing window and 20 requests per minute. Full access to the archive, structured salary data and higher limits is available for products and research: email marc@startup.jobs.
Can startups post jobs through an agent?
Not yet. Posting a job through MCP, a CLI and webhooks for new listings are on our roadmap. If you would use one of them, email marc@startup.jobs and tell us what you are building; it decides what we ship first.
How is this different from scraping LinkedIn or Indeed?
You get one hosted endpoint with de-duplicated startup listings, company context and market tools, not a scraper to run and maintain. Startups post directly on startup.jobs, and we also collect openings from their own careers pages, so the data is current and startup-only.
Give your agent startup jobs
One endpoint, no account, about a minute to set up.