PreHireBadge
Claudeまたは任意のMCPクライアントから、FCRA準拠の採用前および業務委託者向けバックグラウンドチェックを注文できます。チェックを依頼(候補者への招待は同意取得が先)、ステータスの確認、候補者一覧の取得、クレジットの確認、Adverse Actionの次の手順の取得が可能です。PKCEを使用したOAuthに対応。記録行や採用判断は提供しません。1件あたり5ドル。
ホスト型 MCP サーバー
npx add-mcp 'https://www.prehirebadge.com/api/mcp'Claude Code、Codex、Cursor などにインストールできます
ドキュメント
Order background checks from Claude.
PreHireBadge exposes an MCP server, so Claude and other MCP clients can request a background check, read its status, list candidates, and check your credit balance — after an owner or admin approves the connection.
Claude · PreHireBadge connected
Run a background check on Sam Rivera for a driver role in Phoenix. Offer made.
request_background_check
{ firstName: "Sam", lastName: "Rivera", positionTitle: "Driver", … }
Invite sent to Sam. I'll be able to read the status with get_check_status once they complete it. I can't make a hiring recommendation.
Illustrative example with sample data.
How it works
4 steps, start to finish
- 1
Add the server
Point your MCP client at PreHireBadge. It discovers OAuth automatically. - 2
Approve the client
Sign in, pick scopes, approve. The browser returns a one-time code. - 3
Ask your assistant
Request a check, read status, or check credits in plain language. - 4
Candidate gets the invite
The same invite email the dashboard sends. Identifiers stay with the candidate.
Background checks as a tool your AI can call
The Model Context Protocol lets AI assistants call tools in other products. PreHireBadge's MCP endpoint is POST /api/mcp, speaking JSON-RPC 2.0, and it uses the same account, credits, and rules as the HTTP API.
Connecting is standard OAuth. The client discovers the authorization server, registers, and opens the approve page with a PKCE challenge. An owner or admin signs in, chooses scopes — checks:read, checks:write, candidates:read, credits:read, reports:read — and approves. Access tokens last one hour.
The tools are request_background_check, get_check_status, list_candidates, interpret_result, and get_credit_balance. The assistant can order an invite and read status, but it never receives record rows, and interpret_result lists the adverse-action steps while refusing any hire or no-hire verdict.
Orders placed over MCP go through the same path as every other channel: the same consent-first invite email, the same fair chance gates for the work location, and the same credit. The candidate still enters their own identifiers, so nothing sensitive passes through the assistant's conversation.
FAQ
Questions employers ask about claude & mcp
Still unsure? Ask Peter, our assistant, in the bottom-right corner, or browse the help center.
Which MCP tools does PreHireBadge provide?
request_background_check, get_check_status, list_candidates, interpret_result, and get_credit_balance.
How does Claude authenticate to PreHireBadge?
OAuth 2 with PKCE (S256). The client registers, an owner or admin approves it in the browser, and the client exchanges the code for a bearer token that lasts one hour.
Can Claude see criminal records through MCP?
No. Record rows are not included in MCP responses. The assistant sees status and summaries.
Will the AI tell me whether to hire someone?
No. interpret_result lists adverse-action steps and refuses a hire or no-hire verdict.
Does ordering through Claude cost more?
No. An MCP order spends the same single report credit as a dashboard or API order, from the same balance.
Does it work with assistants other than Claude?
Yes. Any MCP client that supports OAuth discovery and PKCE can connect. It finds the authorization server at /.well-known/oauth-authorization-server and registers itself.
Your next candidate is one report away.
No credit card to sign up, no expiration on what you buy. Run one report for $10, or stock up with the 10-pack for $90 and use them whenever hiring picks up.
- No monthly fees
- FCRA compliant
- Most clear reports same day. Records verified within 72 hours.
- No contracts