TypeScript framework for AI agents
Developed by the Astro team, Flue stands out as a TypeScript-based sandbox agent framework. This structure allows developers to create artificial intelligence agents in secure and isolated environments.
Updates
- September 19, 2026: Stars 8,255 → 8,295, latest release @flue/cli@2.1.0 (September 18, 2026).
- September 17, 2026: Stars 8,244 → 8,255, latest release @flue/cli@2.0.8 (September 16, 2026).
- September 16, 2026: Stars 7,625 → 8,244, latest release @flue/cli@2.0.7 (September 15, 2026).
- August 2, 2026: Stars 4,594 → 7,625.
What you get
- Creating programmable and headless agents based on TypeScript.
- Fast and scalable working environment with virtual sandbox.
- Versatile deployment across Node.js, Cloudflare, and CI/CD processes.
Installation
flue dev --target nodeflue build --target node # Node.js server (single bundled .mjs)
flue build --target cloudflare # Cloudflare Workers + Durable ObjectsRunning it
flue run hello --target node \
--payload '{"text": "Hello world", "language": "French"}'If you don't write code
I want to develop an artificial intelligence agent using the Flue framework. How can I define a workflow using TypeScript in my project? Specifically, how can I configure the model with the createAgent function and interact with my agent with session.prompt? Using a simple 'hello-world' example, can you explain step by step how I can start an agent at runtime and get results?
Related dictionary terms
Links
TreScout did not build this tool · we found it in GitHub trends and wrote it up. This page describes the repository as of 2026-06-06: The star count and our text belong to that day, the repository may have changed since. Check the repository link for the current state. This page was machine-translated from the Turkish original · the Turkish version prevails.