UnifAI Tools MCP

Example Use Cases

• 1

  Discover and call external tools dynamically based on semantic search rather than exposing all tools statically.

• 2

  Expose a fixed set of toolkits or individual actions to MCP clients like Claude Desktop for consistent, pre-defined capabilities.

• 3

  Seamless integration of tool results into OpenAI-compatible function-calling workflows so LLMs can iterate with live tool outputs.

Description

This project provides a Python SDK and an MCP server integration for UnifAI, enabling agents to discover, search, and invoke tools (actions and toolkits) dynamically or via static exposure. It supports creating toolkits, registering action handlers, and running tool servers that MCP clients (e.g., Claude Desktop) can consume using a simple uv-based launcher. The server streamlines tool-driven agent workflows and integrates with OpenAI-compatible function-calling LLMs so tool calls can be executed and returned to the model. API keys are available for free from UnifAI to use agent and toolkit functionality.