Fast Agent is an open-source Python framework for building, running, and evaluating AI agents, published on GitHub where it has gathered about 3,944 stars and is under active development in 2026. It aims to make agent development feel as productive as modern web frameworks: define agents in simple configuration, connect them to models and tools, and run them interactively from the terminal or as part of automated evaluations.
The framework’s standout strength is protocol support. Fast Agent works as both an MCP client and MCP server, and also speaks ACP and A2A, so agents can use the fast-growing ecosystem of MCP tool servers, expose themselves as tools to other systems, and interoperate with other agents. Declarative YAML configuration, a rich interactive CLI, multi-model support across providers, human-in-the-loop workflows, and agent skills round out a toolkit designed for people who iterate on agents daily and need to measure whether changes actually improve them. Chaining and orchestrating multiple agents is supported out of the box, including patterns where agents hand work to one another or run side by side for comparison — the kind of workflow the framework’s name promises to keep fast.
Fast Agent is free and open source under the Apache-2.0 license. Install it from PyPI and build agents at no cost; you only pay for any model APIs or MCP services you choose to connect.
Fast Agent is for developers and agent builders who live in the MCP ecosystem and want a fast, code-friendly way to prototype, connect, and evaluate agents across models and tools without heavy framework overhead.