Connect these docs to Claude, ChatGPT, GitHub Copilot, or any other AI tool. Your assistant can then look up Checkr's real endpoints and operations while it works.
For per-page AI actions inside the docs themselves, see Ask AI.
These docs run a Model Context Protocol (MCP) server that gives an AI agent live, structured access to every page and every API operation:
https://docs.checkr.com/mcpIt speaks streamable HTTP and needs no credentials — it only serves published documentation.
This is not the Checkr product MCP. The Docs MCP server serves documentation, and nothing else. It cannot read candidates, reports, or any account data.
To give an agent access to real background check data, use Checkr's product MCP servers at mcp.checkr.com instead — see What is MCP?.
Most clients discover the server automatically; to check it's reachable yourself, curl its card: https://docs.checkr.com/.well-known/mcp/server-card.json.
Run:
claude mcp add --transport http checkr-docs https://docs.checkr.com/mcpThen run /mcp to confirm the server is connected.
You can name the server anything; checkr-docs is used throughout this page to keep it distinct from the checkr product MCP servers.
| Capability | Use it for |
|---|---|
search | Full-text search across every docs page |
listApis | Discover which APIs are published and what each covers |
getEndpoints | List every endpoint on a given API |
getEndpointInfo | Parameters, request and response schemas, and examples for one endpoint |
getSecuritySchemes | How authentication works for a given API |
getFullApiDescription | Retrieve the complete OpenAPI description |
Your client may show only two tools, describe-tools and execute. That is the server's calling convention — the capabilities above are reached through them, and the agent handles this for you.
Once connected, ask in plain language and name Checkr so the agent goes to the server:
- "Using the Checkr docs, how do I get a staging API key and authenticate my first request?"
- "Using the Checkr docs, walk me through running my first background check."
- "Using the Checkr docs, what information do I need to collect from a candidate before I can order a report?"
- "Using the Checkr docs, how do I get notified when a report finishes? Write me a handler for it."
Every page on this site has a Markdown twin: append .md to the URL.
curl https://docs.checkr.com/get-started/quick-start.mdThis works for generated API reference pages too, which makes it a precise way to ground a single endpoint:
curl https://docs.checkr.com/apis/openapi/candidates/createcandidate.mdFrom inside the docs, the page actions menu at the top of every page does the same thing in one click — Copy page, Open in ChatGPT, or Open in Claude. See Ask AI for the full menu.
These docs also publish llms.txt, an index of every page's Markdown URL. The MCP server above is our recommended way to retrieve pages on demand.
Checkr's OpenAPI description has exact request and response shapes, required fields, enums, and error codes. Give it to your assistant as grounding context.
- With the Docs MCP server connected, ask for it by name — the agent calls
getFullApiDescription, orgetEndpointInfofor a single operation. - Without MCP, pull the Markdown twin of the specific operation pages you care about, as shown above.
The full spec is large. For focused questions, give the assistant just the one or two operations you're working on. You'll get better answers and use far fewer tokens.
Never share production API keys with AI tools. Use staging credentials for all AI-assisted development.
- Prefer the MCP server over pasted copies. Pasted docs go stale the moment they're pasted; the server always serves what's published now.
- Use the staging environment. All AI-assisted testing should target
api.checkr-staging.com. - Say you want the Checkr docs. Describe the task in your own words — "using the Checkr docs, how do I start a background check?" — and let the agent find the right endpoints. Mentioning Checkr and the docs is usually enough to point the agent at the server rather than its own memory, and it keeps the request clear if you also have Checkr's product MCP servers connected.
- Review generated code before deploying. AI can produce plausible-looking but incorrect code — always verify against the API reference.