Show HN: Toolnexus for Python – MCP, agent skills,a2a for any LLM
Toolnexus is a Python framework that enables integration of MCP, agent skills, and A2A protocols for use with any large language model.
Background
- **Toolnexus** is a new Python package (by the same creator as the popular `smolagents` library) that aims to unify how AI agents connect to external tools. Instead of coding separate integrations for each AI model, it provides a single connector that works across APIs like OpenAI, Anthropic, Google Gemini, and open-source models via Ollama.
- The project supports three key protocols: **MCP** (a standard for tool-calling first defined by Anthropic), **A2A** (Agent-to-Agent — Google's emerging standard for agents talking to each other), and custom "agent skills" (reusable tool modules from the `smolagents` ecosystem).
- This matters because the AI tool ecosystem is currently fragmented — every model provider has its own way of letting agents call functions, search the web, or run code. Toolnexus lets developers write tools once and use them with any LLM, similar to how SQLAlchemy abstracts different databases behind one interface.
- The GitHub repo is a single-file source (< 1,000 lines) that's BSD-licensed, making it easy to audit or fork.