@postman/postman-mcp-server

What it does

The server connects Claude and other AI agents to the Postman API, exposing programmatic access to workspaces, collections, and environments. It offers three configurable tool sets—Minimal (essential operations), Code (API client code generation), and Full (comprehensive API tooling)—and runs either as a managed remote service with OAuth or as a local process authenticated with an API key.

Who it's for

API developers and teams integrating AI into their development workflow. Backend engineers who need to query Postman collections during coding or CI/CD setup. QA and automation engineers automating API testing and validation at scale.

Common use cases

• Query collections and environments from Claude Code to surface API schemas and endpoints while debugging or writing integration tests
• Generate client code in multiple languages from OpenAPI definitions stored in Postman
• Automate collection exports, environment variable synchronization, and workspace management through natural language prompts
• Extract API metadata (schemas, auth schemes, endpoints) to populate documentation sites or developer dashboards
• Run the server locally to keep sensitive API definitions off third-party infrastructure

Setup pitfalls

• The remote server (https://mcp.postman.com) authenticates with OAuth, while the local server requires a Postman API key in the POSTMAN_API_KEY environment variable—do not assume both support the same authentication method
• The tool set (Minimal, Code, Full) determines which capabilities are available; choose the right mode upfront to avoid mid-workflow discovery that a tool doesn't exist
• The server reads and writes collections to the filesystem; isolate its working directory if the host runs untrusted code
• OAuth setup and configuration differ by IDE and MCP host; check Postman's integration guide for your specific tool to avoid misconfiguration