any-agent

Active
GitHub Python Apache-2.0

Description

A single interface to use and evaluate different agent frameworks.

Key Features

  • Unified interface supporting 7+ major agent frameworks: TinyAgent, Google ADK, LangChain, LlamaIndex, OpenAI Agents, Smolagents, Agno
  • Built-in agent evaluation capabilities to compare performance and output quality across frameworks
  • Supports MCP (Model Context Protocol) tool integration and A2A (Agent-to-Agent) communication
  • Built-in tracing and serving functionality for monitoring and deploying agent services
  • Rich cookbook examples: first agent, evaluation, callbacks, MCP integration, A2A deployment and more
  • Supports Agents-As-Tools pattern for building multi-agent systems

Use Cases

πŸ’‘ Quickly compare output quality and performance across different agent frameworks (LangChain vs OpenAI Agents vs Smolagents)
πŸ’‘ Build framework-switchable agent applications using a unified API, avoiding framework lock-in
πŸ’‘ Integrate external tools via MCP to build agents with web search and information retrieval capabilities
πŸ’‘ Deploy multi-agent collaboration systems using the A2A protocol
πŸ’‘ Use built-in evaluation tools for automated quality detection of agent outputs

Tags

a2a agent-evaluation agents ai mcp mcp-tools observability python

Categories

πŸ”Œ MCP Tools πŸ“Š Observability

Quick Start

Install: pip install any-agent. Set API Key (export MISTRAL_API_KEY=YOUR_KEY). Create agent using unified interface: from any_agent import AgentConfig, AnyAgent, call AnyAgent.create("tinyagent", AgentConfig(...)) to create, then agent.run(query) to execute.
Visit GitHub Visit Website

Related Projects