fast-agent

Active
GitHub Python Apache-2.0

Description

Code, Build and Evaluate agents - excellent Model and Skills/MCP/ACP Support.

Key Features

  • CLI-first interactive agent framework with prompt_toolkit-powered TUI, shell completions, and streaming responses via Rich
  • Complete MCP feature support including Sampling and Elicitations — the first framework with end-to-end tested MCP integration
  • Declarative agent definition with minimal boilerplate — define agents via @fast.agent decorator and compose prompts with MCP servers
  • Comprehensive model support — native Anthropic, OpenAI, Google, plus Azure, Ollama, Deepseek, TensorZero, and local llama.cpp
  • Agent Skills (SKILL.md), MCP-UI support, OpenAI Apps SDK (Skybridge), and ACP client compatibility
  • Built-in MCP transport diagnostics with OAuth KeyRing storage for inspecting Streamable HTTP usage

Use Cases

💡 Building and testing multi-agent workflows with MCP server orchestration and human-in-the-loop feedback
💡 Rapid prototyping of LLM-powered tools using declarative agent definitions and hot-swappable models
💡 Evaluating model performance across providers by switching models with a single --model flag
💡 Developing coding agents that leverage shell mode, skills, and MCP tool chains for autonomous tasks
💡 Creating production agent applications with structured outputs, PDF/vision support, and persistence

Tags

agent-framework mcp skills python

Categories

🤖 Agent Framework 🔌 MCP Tools

Quick Start

Install uv, then run `uvx fast-agent-mcp@latest -x` for an interactive session. Or install permanently with `uv tool install -U fast-agent-mcp` and start with `fast-agent go`. Define agents in Python with `@fast.agent(instruction="...")` and run via `uv run agent.py`.
Visit GitHub

Related Projects