Get started
MCP for AI agents
Connect Claude or Cursor to the DKBSign API documentation through the gateway’s MCP server—same OpenAPI spec as this site, without calling the live API for you.
How it works
The DKBSign API Gateway hosts a Model Context Protocol (MCP) server at https://api.dkbsigns.com/mcp. When you connect Claude or Cursor to that URL, the assistant can read the official API description (paths, methods, request bodies, responses, and auth notes)—the same OpenAPI document that powers this documentation site.
Transport is streamable HTTP: the client sends POST requests to /mcp on the gateway host. The server does not execute API calls on your behalf; it only returns documentation. You still integrate with Quickstart and Authentication when building real clients.
What you get
The connector exposes an Api Documentation Server with:
- Search — find endpoints by keyword (path fragment, tag, summary, etc.).
- Endpoint details — full schema for a path and HTTP method.
- Full OpenAPI — entire spec as a resource (best for broad questions; search + details are usually enough).
Claude
Claude connects to remote MCP servers as a custom connector (not via the local claude_desktop_config.json stdio entries).
- Open Settings → Connectors → Add custom connector (choose Web if prompted).
- Set Remote MCP server URL to
https://api.dkbsigns.com/mcp. - Under authentication, choose No sign-in unless your organization has protected
/mcp. - Save, then enable the connector in the chat where you want API documentation tools available.
On claude.ai, add the connector under organization or personal settings the same way, then turn it on per conversation.
Cursor
- Open Cursor → Settings → MCP.
- Add an HTTP (remote) server.
- Set the URL to
https://api.dkbsigns.com/mcp(or your tunnel URL in local dev). - Restart or refresh MCP if Cursor does not list the tools immediately.
Cursor must be able to reach the URL from your machine (unlike Claude cloud connectors, which call the server from Anthropic's network). For local gateway dev, use a published port or tunnel; for production, use https://api.dkbsigns.com/mcp.
Using it in a chat
With the connector enabled, ask the model explicitly to use the API documentation MCP. A useful pattern:
Use the DKBSign API documentation MCP.
1. Search for endpoints related to "login".
2. Get full details for POST /api/v1/login (request body and responses, including MFA).
3. Search for "envelopes" and summarize the main create/send endpoints with example JSON from the spec.For implementation help, ask for curl or code samples from the spec after fetching endpoint details. Remind the model that protected routes need POST /api/v1/login and Authorization: Bearer {access_token}.
If answers look outdated, the published OpenAPI on the gateway may need refreshing—check the live reference on API reference or https://api.dkbsigns.com/docs.