We’ve just released MCP Kit (https://github.com/agentiqs/mcp-kit-python), an open-source Python toolkit to help developers build, test, and connect AI agents to real-world tools, APIs, and services using the Model Context Protocol (MCP).
We built this at Agentiqs (https://agentiqs.ai) while experimenting with multi-agent workflows. We found ourselves repeatedly writing glue code to connect OpenAPI specs, mock APIs, and deploy custom toolchains—so we decided to package that all into a unified, config-driven toolkit.
What it does:
* Convert REST APIs into MCP-compatible tools using OpenAPI specs.
* Mock tool behavior using either LLMs or random generators for development.
* Combine multiple tools/services into one interface with multiplexing.
* Use the same targets across OpenAI Agents, LangGraph, and raw MCP sessions.
* Configure workflows using YAML or JSON, and deploy anywhere.
Example: Want to simulate a tool with realistic behavior before building the backend? Just point MCP Kit at a spec, add a mock LLM config, and it works out-of-the-box—no scaffolding needed.
Try it now:
uv add mcp-kit
Or check out the examples: https://github.com/agentiqs/mcp-kit-python/tree/main/example...
Why we built it:
We’re building a platform that heavily relies on connecting agent tools across many teams and APIs. There wasn’t a lightweight, flexible, and modular solution that spoke MCP fluently and supported both real and mocked workflows for testing. So we built one.
It’s still early, but stable and usable—we’d love feedback from anyone working with AI agents, developer tools, or protocol-based integrations. Happy to answer questions or discuss implementation details.
GitHub: https://github.com/agentiqs/mcp-kit-python
Full docs and examples: https://agentiqs.ai/docs/category/python-sdk
PyPI: https://pypi.org/project/mcp-kit/
Thanks! Mauro & Guido @ Agentiqs