How can we help?

A match is produced from three inputs: the job ad text, your profile claims, and your stated constraints (location, salary floor, visa).

What is read from the ad

Title, seniority, required skills, and any hard filters the employer states — years of experience, clearance, right to work.

What is read from your profile

Each claim you have added, with its evidence. A claim with no evidence still counts, but it is flagged before it goes to an employer.

Why a score can be low

Most low scores come from a hard filter, not from a weak profile. The reason is always shown next to the score.

A claim is one statement a reader can check. "Led the migration of 40 services to Kubernetes" is a claim. "Extensive DevOps experience" is not.

Rules

• One fact per claim. Split anything with an "and" in it.
• Attach the evidence — repository, ticket, publication, reference.
• Numbers beat adjectives. Scope, volume, duration.
• Never claim what you cannot show. A flagged claim costs more than a missing one.

The MCP server exposes your job search as tools any MCP-capable agent can call.

What it exposes

• Search live roles with the same filters as the web app.
• Save a role, and read your saved list.
• Read and update your application tracker.

Connecting

Authorisation is a one-time browser consent. Your agent never sees your password.

The agent reports an authentication error, or returns no tools at all. Almost always this is the authorisation step never completing, rather than the server being unavailable.

1. Confirm you are signed in to HiredSignal in a browser on the same machine.
2. Start the connection from your agent again. A browser window should open asking you to authorise it — if none opens, check that your pop-up blocker is not suppressing it.
3. Approve the request. The agent should report the tool list within a few seconds.
4. If it still fails, revoke the agent's access in your account and authorise it once more, so a stale token is not being reused.
5. If the tool list appears but every call returns an error, check your plan — an expired subscription leaves the connection working and the calls refused.

The account exists but cannot be used until the address is confirmed.

1. Check the spam or junk folder — verification mail is the message most often filtered.
2. Confirm the address you typed, including the domain. A typo produces an account that can never be verified.
3. Request the verification mail again from the sign-in page.
4. If nothing arrives within fifteen minutes, contact support with the address you used and we will confirm whether it was accepted.

Job data is collected from public postings, so coverage varies by market and by how the employer publishes.

1. Widen the terms first — a single very specific title matches far fewer postings than the same role described two ways.
2. Remove location filters and re-run. Remote and hybrid postings are frequently filed under a head-office location.
3. Check whether you are at your plan's request quota; once it is reached, further calls are refused rather than returned empty.
4. If a specific employer is missing entirely, tell us who — that is a collection gap and it is fixable.

Payment and entitlement are separate records, and the second one follows the first. Usually this resolves itself within a minute.

1. Sign out and back in, so the session picks up the new entitlement.
2. Check the email receipt from Stripe — if there is no receipt, the payment did not complete.
3. If the receipt exists and the limits have not changed after five minutes, contact support with the receipt and we will reconcile it by hand.

It is an endpoint that lets an AI agent search HiredSignal on your behalf, using your account and your quota. You do not need it — the web application works on its own — but it is the point of the product if you already work through an agent.

Any client that speaks the Model Context Protocol and supports OAuth authorisation. You authorise it once in a browser; the agent holds a token after that.

Yes. Revoking a token stops that agent immediately and leaves your account and any other agents untouched.

Publicly posted advertisements on employer sites and job boards. They are provided as they were found — a vacancy can be filled or withdrawn between collection and your seeing it.

Not unless you explicitly turn that on. By default HiredSignal searches and returns listings; it does not contact employers.

No. See the privacy policy — your CV and your searches are not used as training data.

Requests are refused with a quota error rather than returning partial results, so an agent cannot silently work from a truncated answer.

From your billing page. Access continues to the end of the period you have already paid for.

Requests authenticate with a bearer token issued to one agent.

Authorization: Bearer <token>

Getting a token

Authorise the agent once in a browser. The agent holds the token afterwards; your password is never seen by it.

Revoking

Revoking a token stops that agent immediately and leaves your account and every other agent untouched.

GET /v1/jobs?q=platform+engineer&location=london&remote=true&limit=20

Parameters

• q — free text matched against title and description.
• location — city or region; omit for everywhere.
• remote — true restricts to remote-eligible postings.
• limit — 1 to 100, default 20.

Rate limits

Requests are counted against your plan's quota. Once it is reached, calls are refused rather than returned empty — an empty result and an exhausted quota must not look alike.

{"error":{"code":"unavailable","message":"content is temporarily unavailable","request_id":"a6c6122d"}}

Codes

• bad_request — the parameters are wrong; the message says which.
• unauthorized — no token, or a revoked one.
• quota_exceeded — the plan's request budget is spent.
• unavailable — an upstream is down. Retry; this is not your fault.