Any AI agent can now search our jobs, through an open standard

If you ask an AI assistant to find you a job today, it mostly scrapes. It loads pages built for human eyes and guesses at what’s a salary and what’s a footer. That’s why the answers are often stale, or quietly wrong.

freehire now supports OJCP — the Open Job Context Protocol, a draft open standard for how an agent asks a job board for jobs directly. An agent reads one file at /.well-known/ojcp.json, learns what we offer, and starts asking. Nobody has to write an integration for freehire specifically — that’s the whole point of a standard.

Three things an agent can do:

• Search the catalogue — the same postings you see on the site, narrowed by location, salary, employment type, seniority and how recently they went up
• Pull one posting in full — description, salary, skills, how to apply
• Ask about an employer — who they are, what they’re hiring for

The standard defines fewer filters than the site has, so the finer ones — category, skills, hybrid versus on-site, the eligibility gates — are not reachable through it yet. Where an agent asks for something OJCP cannot express, the answer says which parameter it had to ignore rather than quietly returning more than was asked for.

It works over plain HTTP or over MCP, the protocol most AI tools already speak. No API key, no signup, nothing to pay — same as our public API. The budget is 300 requests a minute per anonymous caller.

We’re one of the first implementers outside the group that drafted it, and we’ve been sending our findings back. Writing a real provider turned up several things the standard can’t express yet — a job open in more than one country, for one — each measured against our 5.9 million open postings. Two of those are now proposed changes to the spec.

Nothing about the site changes. This is for the agents.