Mcpgen – Turn OpenAPI/Postman Specs into Python MCP Servers
Generates readable, editable Python code you own instead of a black-box runtime proxy.
Understand any enterprise platform in minutes. Local CLI, no login.
MCP server feeds OpenAPI ground truth to agents, stopping them from hallucinating endpoints.
Backend engineers, Integration developers, AI agent users
Cursor · Sourcegraph Cody · Postman
I've been building Scout over the past few months.
Point it at an OpenAPI spec and documentation, and it builds a grounded understanding of the platform: architecture, entities, workflows, starter code, a threaded AI chat, and an MCP server for coding agents.
Everything runs locally. No accounts or hosted backend.
One of the problems I kept running into was coding agents confidently inventing endpoints because they only knew APIs from training data. Scout uses the OpenAPI spec and docs as ground truth instead.
It's MIT licensed and I'd love feedback from people building integrations or using coding agents heavily.
GitHub: https://github.com/prabhuavula7/scout ...
npm: npm i -g @dotapk7/scoutcli
Generates readable, editable Python code you own instead of a black-box runtime proxy.
Native binary output beats curl wrappers when you need typed API commands.
No-code OpenAPI import beats writing custom MCP servers for each API.
MCP server per project lets AI agents query your API spec inside Cursor.
Auto-generates typed MCP tools from OpenAPI or DB schemas without custom code.
Progressive disclosure beats MCP tool flooding for large OpenAPI specs.