Like many of you, we struggled with sending massive, messy JSON responses from existing APIs to LLMs. It wastes tokens, increases latency, and honestly, most models get lost in the noise.

That’s why we built reShapr: an open source, no-code MCP Server that acts as an intelligent proxy between your LLM and your APIs (REST, GraphQL, gRPC).

The Goal: Reshape complex API payloads into lean, structured context before it hits the model.

🛠 Why reShapr?

• No-Code & Schema-First: Point to your backend API using your OpenAPI/Swagger, GraphQL schema, or gRPC protobuf file.
• Context Control: Filter and transform data on the fly to keep prompts surgical.
• MCP Native: Works out of the box with Claude Desktop or any MCP-compliant client.

​❤️ From the Community, For the Community

We are firm believers in the power of open source. reShapr was launched by the maintainers of Microcks, a CNCF incubating project. Just like with Microcks, we are fully community-driven and committed to the "open" in open source.

I'm posting here to grab feedback. If you like the vision, please join the community and help us shape the future of API-to-LLM integration!

Get Started:

• Quick-start (Docker): https://reshapr.io/docs/how-to-guides/docker-compose
• Try it Online: https://reshapr.io/docs/tutorials/try-reshapr-online
• GitHub: https://github.com/reshaprio/reshapr

We want to hear from you! How are you handling API-to-AI context today? Would love to hear your thoughts, feature requests, or even your "nightmare JSON" stories.