Skip to content

Quickstart

Import an API, connect credentials, publish an MCP endpoint and call your first tool.

This guide takes you from an API definition to an AI client calling your API through PipMCP. You need:

  • A PipMCP account (create one free).
  • An API definition: an OpenAPI 3.x or Swagger 2.0 file or URL, a Postman collection, a GraphQL schema, or a documentation URL.
  • A credential for that API (API key, bearer token, basic auth or OAuth 2.0 client), unless the API is public.

No API handy? Open Integrations and choose Try the demo API. It imports a built-in sample API with customers and invoices, and demo calls do not count toward your usage.

1. Add your API

In the dashboard open Integrations and choose Add API. Paste a spec URL such as https://api.example.com/openapi.json, upload a file, or paste the definition.

PipMCP fetches and parses the definition, maps each endpoint to a business action, and generates tools. Progress is shown live.

Details: Import an OpenAPI spec.

2. Review the generated tools

When the import finishes you land on the review screen:

  • Tools lists everything that was generated, for example search_customers, create_invoice or delete_customer, with a risk level for each.
  • Anything PipMCP was not confident about is marked Needs review. Confirm or correct those mappings; publishing is blocked until they are resolved.
  • Disable any tool you do not want agents to use.

Details: Tools and mappings.

3. Connect credentials

Open Connections and add a connection for the integration. PipMCP pre-selects the auth type it found in your spec. Enter the secret and choose Test connection.

Secrets are encrypted immediately and are never shown again. Details: Connect credentials.

4. Publish an MCP endpoint

Open MCP Endpoints, choose Create endpoint, pick the integration and connection, and choose a permission preset. Safe writes is a good default: reads and ordinary writes run, destructive and financial actions ask for confirmation.

Create an access token. Copy it now: it starts with pmcp_ and is shown only once.

Your endpoint URL looks like this:

https://mcp.pipmcp.com/<workspace>/<endpoint>

Details: Publish an MCP endpoint.

5. Connect an AI client

For Claude Code:

claude mcp add --transport http my_api https://mcp.pipmcp.com/<workspace>/<endpoint> \
  --header "Authorization: Bearer pmcp_your_token"

The endpoint page has a step-by-step guide for Claude, ChatGPT and Cursor that creates a personal link (URL with the key built in) for you. See Connect from Claude and Connect a generic MCP client.

6. Watch it work

Ask your agent something like "find the customer Acme and show their open invoices". Each tool call appears in Executions with the tool, client, redacted input, policy decision, duration and result. See Logs.

Next steps