One query, every ATS-hosted job, with what changed since yesterday.

Start

First call, three ways

No SDK required. JSON in, JSON out, cursor pagination, RFC 7807 errors. /stats and /schema are free and need no key.

# Every engineering role opened in the last 24 hours
curl -H "Authorization: Bearer $ROLESET_KEY" \
  "https://api.roleset.io/v1/jobs/changes?event=opened&department=Engineering"

# Remote backend roles in Germany
curl -H "Authorization: Bearer $ROLESET_KEY" \
  "https://api.roleset.io/v1/jobs/search?q=backend&remote=true&country=DE"

# Is this company growing, and where?
curl -H "Authorization: Bearer $ROLESET_KEY" \
  "https://api.roleset.io/v1/companies/acme.com/hiring-signal"

import httpx

api = httpx.Client(
    base_url="https://api.roleset.io/v1",
    headers={"Authorization": f"Bearer {KEY}"},
)

# Walk the change feed. Keep the cursor; resume from it next run.
cursor = None
while True:
    page = api.get("/jobs/changes", params={
        "since": "2026-09-05T00:00:00Z",
        "cursor": cursor,
        "limit": 100,
    }).json()

    for event in page["data"]:
        print(event["event"], event["job"]["title"], event["diff"])

    cursor = page["meta"]["next_cursor"]
    if not cursor:
        break

// Five tools: search_jobs, job_changes, get_job,
// company_jobs, hiring_signal
{
  "mcpServers": {
    "roleset": {
      "type": "http",
      "url": "https://mcp.roleset.io",
      "headers": { "Authorization": "Bearer YOUR_KEY" }
    }
  }
}

# Claude Code
claude mcp add --transport http roleset https://mcp.roleset.io \
  --header "Authorization: Bearer $ROLESET_KEY"

Endpoints

Seven, and one of them is the point

/jobs/changes is what separates this from a list of feed URLs. Poll it instead of re-running a search and diffing the results yourself.

The record

One schema across nine platforms

Nine ATS platforms disagree about almost everything. Normalizing that disagreement is the work you are paying not to do.

Freshness

Every active board, inside 24 hours

And we tell you when. Every response carries meta.last_crawled_min — the oldest successful crawl behind the exact rows you were handed, so you can caveat an answer instead of guessing at it.

Pricing

Pay for what you pull

A record is one job posting, or one change event, returned to you. /stats and /schema are free and need no key.

Prices exclude sales tax and VAT. Paddle is the merchant of record and adds them at checkout where they apply, so a US or EU card sees a little more than the figure above.

What is not in here

The index has edges, and they are on purpose

• No personal data. No applicants, no recruiter names, no contact details. Email addresses and phone numbers are stripped from descriptions before anything is written down.
• No aggregators. Official public ATS endpoints only. No LinkedIn, no Indeed, no HTML scraping.
• No inferred fields. Seniority and skills are not guessed at. If a board does not publish a salary, the salary is null.
• A posting can close between crawls. Freshness is a 24-hour guarantee, not a live socket. last_crawled_min tells you where you stand.