FastAgency

What is FastAgency?

FastAgency is an open-source Python framework for moving multi-agent workflows from prototypes into production. Rather than defining a new agent runtime, it provides a unified deployment interface around workflows written in supported agent frameworks. The analyzed documentation states that the current runtime is AG2, formerly AutoGen.

A workflow can be reused across console and Mesop web interfaces, exposed through a FastAPI REST service, or distributed across workers with a NATS message broker. Developers can import an OpenAPI specification to give agents external API capabilities. A Tester class supports automated workflow checks in continuous integration, while the command-line interface runs and monitors applications from a terminal.

How to Use FastAgency

1. Create a Python project using the recommended project template or a virtual environment.
2. Define an AG2 multi-agent workflow and its inputs, interactions, and output.
3. Run and debug the workflow locally with a console or Mesop interface.
4. Add automated tests with the Tester class and run them in continuous integration.
5. Select a deployment adapter for Mesop, FastAPI, or NATS-based distributed workers.
6. Build and deploy the generated application files to the chosen infrastructure.

Core Features

• Unified programming interface: Reuses one workflow across command-line and web user interfaces.
• AG2 runtime: Runs multi-agent workflows created with AG2, formerly AutoGen.
• External API integration: Imports OpenAPI definitions and connects operations to agents with minimal code.
• Workflow testing: Automates agent interaction checks through a CI-friendly Tester class.
• Deployment adapters: Serves workflows with FastAPI or coordinates distributed workers through NATS.
• Command-line orchestration: Runs workflows, passes parameters, and monitors interactions from a terminal.

Use Cases

• Prototype deployment: Convert a notebook-based AG2 workflow into a production application.
• Agent web apps: Add an interactive Mesop chat interface to a multi-agent system.
• Agent APIs: Expose workflows through a scalable FastAPI service.
• Distributed execution: Coordinate workers across machines or data centers through a message broker.
• Reliable releases: Test agent behavior before deploying workflow changes.

Pricing

FastAgency is presented as an open-source project. Users supply the infrastructure and credentials for their selected language-model provider, so provider and hosting costs are separate.

Frequently Asked Questions

Is FastAgency an agent framework?

It describes itself as a production and deployment layer for workflows written in agent frameworks rather than another agent runtime.

Which agent runtime is currently supported?

The analyzed documentation lists AG2, formerly AutoGen, as the current runtime.

Can FastAgency scale beyond one process?

Yes. It supports FastAPI workers and a NATS-based architecture for distributing work across multiple machines.