About Fastagency
FastAgency is an open source framework for taking multi agent AI workflows from prototype to production. It is explicit that it is not another agent framework. Instead it wraps workflows built with AG2, the project formerly called AutoGen, and gives them a deployment surface, so the same workflow can run as a terminal session, a web chat application or a network service by changing which adapter is attached.
It is built for Python teams who already have agents working in AG2 and now need other people or systems to reach them. A console interface handles command line interaction and a Mesop interface provides a browser based chat application. A FastAPI adapter serves a workflow as a REST API, and a NATS adapter built on FastStream routes work through a message broker for distributed or higher volume setups. New projects are generated from a Cookiecutter template that installs dependencies and sets up a development container for VS Code, so the starting point is a working application rather than a blank file. AG2 is currently the only supported runtime, with support for additional frameworks described as planned.
The framework is free. There is no pricing page, no hosted tier and no commercial edition; the code and documentation are published under the ag2ai organization on GitHub. Costs come from whichever model provider the underlying agents call.
One thing to check before adopting it: the documentation site serves a notice that the page being viewed is not the latest version, and the documentation home carries a date in April 2025, so confirm the current release state in the repository.