# toolnexus

> A small, vendor-neutral library that gives any LLM dynamic tool-calling: an MCP host, agent skills, native + HTTP + built-in tools, remote A2A agents, a human-in-the-loop suspension layer, and a unified client loop — byte-identical across JavaScript, Python, Go, Java, C#, Elixir and Clojure — all seven published to their registries.

toolnexus unifies every tool source (MCP servers, agent skills, your own functions via defineTool, HTTP/REST endpoints, built-in shell/file tools, and remote A2A agents) behind one Tool interface, emits schema in OpenAI/Anthropic/Gemini formats, and ships a client with a built-in tool-calling loop (skills injection, parallel + chained calls, hooks, streaming, retries, conversation memory, observability). The same examples/ fixtures produce identical behavior in all seven ports, each held to the full tier by a shared conformance gate. The Clojure port runs one .cljc source tree on both the JVM and cljgo, published to Clojars.

## Documentation Sets

- [Abridged documentation](https://muthuishere.github.io/toolnexus/llms-small.txt): a compact version of the documentation for toolnexus, with non-essential content removed
- [Complete documentation](https://muthuishere.github.io/toolnexus/llms-full.txt): the full documentation for toolnexus

## Notes

- The complete documentation includes all content from the official documentation
- The content is automatically generated from the same source as the official documentation