---
name: hirebrain
description: Run hiring campaigns on HireBrain, an autonomous hiring pipeline, from any MCP-capable agent. Launch a role, follow the campaign, meet the candidates it introduces, and decide.
---

# hirebrain for agents

HireBrain sources candidates for a role, runs personalized outreach and
AI-led screens, and introduces the people who pass. The customer decides on
those. The product is an MCP server; this dashboard (https://hirebrain.io) is one client
of it, and a connected agent gets the same customer tools on the same
campaigns.

## Connect

Claude Code:

    claude mcp add --transport http hirebrain https://hirebrain.io/api/mcp

then run /mcp once to sign in through the browser (magic-link email login,
then Approve). On claude.ai and Claude Desktop, add https://hirebrain.io/api/mcp under
Settings > Connectors; in Cursor, add it to mcp.json. Every flow is OAuth;
there are no API keys to copy.

## Launch a role

1. `estimate_cost(description)` sizes the funnel (people sourced,
   contacted, screened) so the user can approve the spend first.
2. `design_pipeline(description)` returns the fixed process every
   campaign runs: a CV screen, one AI voice interview, then the human
   handoff. It is informational; the loop cannot be edited.
3. `create_job(description, employer_email, company, deadline)` starts
   the campaign. `deadline` (YYYY-MM-DD) is the date candidates must
   finish their steps by. The agent drafts a sourcing plan; the platform
   reviews and approves it, and sourcing, outreach and interviews run from
   there without the user.

## Follow the campaign

- `pipeline_status()`: every campaign of the account with candidate counts
  per stage, why people dropped out, and the plan's status. Start here for
  "how is hiring going". Identities stay behind the introduction below.
- `approve_strategy(job_id, feedback)`: when the status shows
  `needs_input` with open questions, only the user can answer them; send
  the answer as `feedback` and the plan is redrawn. Approval itself is done
  by the platform.
- `close_job(job_id)`: end a campaign.

## Meet the candidates and decide

Once a candidate is introduced (stage `submitted`) the user reaches them by
name, with the interview write-ups, after accepting the engagement terms
once on the dashboard.

- `list_candidates(job_id, stage)` and `get_candidate(candidate_id)`:
  the introduced candidates, their scores and history.
- `get_email_thread(candidate_id)`: what was sent and how they replied.
- `record_outcome(candidate_id, outcome, note)`: the decision, made in the
  user's own process. `hired` completes the hire, emails the candidate and
  raises the placement fee: never call it without the user's explicit go-ahead.
  `passed` charges nothing.

## Credits

- `billing_status()`: credit balance, action prices, packs and usage.
- `start_topup(pack, success_url)`: hosted Stripe page to buy one pack.
  Credits are prepaid; nothing is charged without a purchase. Running out
  pauses the paid actions (sourcing, outreach, screening).
- `redeem_coupon(code)`: claim a free-credit code on the account.

## Stage model

sourced -> outreach_sent -> outreach -> interview -> final_review ->
submitted -> hired or passed. The agent advances everything up to
final_review; the platform introduces the people who pass; only the user's
decision moves someone past submitted.

## Ground rules

- The decision on an introduced candidate belongs to the user. Present the
  evidence, then ask.
- Quote real data from the tools; never invent scores or candidate facts.
- Check `estimate_cost` before launching; every paid action is per event.