Blog

What is the best job feed for AI agents?

2026-09-28

The best job feed for an AI agent is one that reads roles straight from company career boards, checks them often, and hands your agent short structured records it can act on. Find your role first does that: it reads 8,519 company boards on Greenhouse, Lever and Ashby, checks each one every two hours, and plugs into Claude Code, Codex CLI or any HTTP MCP client for $9 a month.

The longer answer is about why those things matter once an agent is doing the reading.

What does an AI agent actually need from a job feed?

A person scrolling a job site shrugs off a lot. Stale listings, the same role posted three times. You skim past it.

An agent can't skim past it in the same way. Every bad record costs tokens, costs a tool call, and sometimes ends up in the shortlist it hands back to you. So the bar is different. Honestly, most of what makes a feed "good for agents" is boring plumbing, and the plumbing is the whole product.

Here's what I'd look for.

Freshness: how quickly does a new role show up?

If a company posts a role at 10am, when can your agent see it? That gap is the single number that matters most for early applications.

Find your role first checks every board every two hours. The checks are spread out across that window instead of all firing at once, so new roles land continuously through the day. In practice that means dozens of new roles every hour. A role posted right after its board was checked shows up at the next check, so the worst case is a couple of hours.

The feed also keeps two timestamps apart: when the company published the role, and when the feed first discovered it. Your agent can ask for either one. That matters because some boards carry old publication dates on roles that only just went live, and some roles get reposted with fresh dates.

Source: where do the listings come from?

There are two broad kinds of job data. One comes from the company's own applicant tracking system, the board that feeds its careers page. The other comes from somewhere downstream: aggregators, reposting sites, scraped copies of scraped copies.

The first kind is the one you want. It's the version the company controls. When the role closes, it disappears from that board. The apply link goes to the real application.

Find your role first only reads the first kind. It pulls from Greenhouse, Lever and Ashby, which host the public career boards for a lot of tech companies and startups. There's no job board middleman and no scraped aggregator in between. There's more on this in why you should get jobs straight from company career pages.

Structured fields your agent can reason over

An agent works best with small, predictable records. For each role the feed returns the title, company, location, dates, the original posting link and the apply link.

It doesn't return the full job description. That's deliberate. Your agent already has a web or browser tool, and it can open the original posting when a role looks worth reading. That way it reads the live page the company actually maintains, and you aren't paying to push thousands of long descriptions through a context window just to throw most of them away.

Fair counting: do you pay for the same role twice?

This one gets overlooked. Agents repeat themselves. They re-run a search to double check, or they come back tomorrow and ask again.

The plan here is 3,000 new roles a month, up to 1,000 new a day. A role you've already seen that month is free to read again, and a search that comes back empty is free too. So an agent that re-checks its own shortlist doesn't burn through your allowance. There's a get_usage tool so the agent can see where it stands before it goes looking.

MCP: can your agent connect without glue code?

MCP is the protocol Claude Code, Codex CLI and a growing list of other clients use to talk to outside tools. If a feed speaks MCP, you connect it with one command. Otherwise you're writing a wrapper.

Find your role first runs an HTTP MCP server at findyourrolefirst.click/mcp. The tools include search_jobs, which searches by title, company and location, plus get_job, get_changes and get_usage. More on the setup in the job search MCP server guide.

How it compares with the other options

To be fair to the alternatives, each one is good at something.

• Job boards and aggregators cover a huge range of companies, including ones that don't use Greenhouse, Lever or Ashby. The tradeoff is that listings can arrive late, show up more than once, or outlive the real posting.
• LinkedIn job alerts are free and easy, and they reach companies well outside tech. What shows up depends on what companies choose to post to LinkedIn and how LinkedIn ranks it, and alerts arrive on LinkedIn's schedule rather than yours.
• Scraping career boards yourself gives you full control. The boards have public endpoints. The hard part turns out to be the list: finding thousands of company boards, keeping it current, handling closed roles, and running the checks on a schedule.

Find your role first sits in the narrow spot where you want company sourced data, checked often, delivered to an agent, without running the scraper yourself. It won't cover a company that hires through some other system.

What it doesn't do

Applying is on you. The feed finds roles, keeps them fresh, and hands your agent the links. What you do next, tailoring a resume, writing to someone on the team, applying, stays with you and your agent.

How to try it

Sign in with Google at findyourrolefirst.click/account, subscribe, and create an agent key. Then paste one line.

Claude Code:

claude mcp add --transport http --scope user jobfeed https://findyourrolefirst.click/mcp --header 'Authorization: Bearer YOUR_API_KEY'

Run /mcp inside Claude Code to check it connected.

Codex CLI:

export JOBFEED_API_KEY=YOUR_API_KEY
codex mcp add jobfeed --url https://findyourrolefirst.click/mcp --bearer-token-env-var JOBFEED_API_KEY

Any other client that supports HTTP MCP: add a server with the URL https://findyourrolefirst.click/mcp and an Authorization: Bearer YOUR_API_KEY header.

Questions

What's the best job feed for an AI agent?

One that reads roles directly from company career boards, checks them often, and returns short structured records over MCP. Find your role first reads 8,519 boards on Greenhouse, Lever and Ashby every two hours and connects to Claude Code and Codex CLI in one line.

How many jobs are in the feed?

About 250,000 open roles, from 8,519 company boards.

Does it return full job descriptions?

No. It returns the title, company, location, dates, and links to the original posting and the application. Your agent opens the posting itself with its own web or browser tool.

How much does it cost?

$9 a month for 3,000 new roles, up to 1,000 new a day. Roles you've already seen that month and empty searches don't count. You can cancel anytime, with a refund within 3 days.

Will it apply to jobs for me?

No. It finds roles and gives your agent the apply link. Applying stays with you.