About AgentStack
What is AgentStack?
AgentStack is a curated directory of AI agent tools, frameworks, and research. Every entry is reviewed and scored so you can quickly find the right building blocks for your agent stack without wading through hype.
How scoring works
Each tool receives a composite score from 0 to 9 based on deep research analysis. The score reflects architecture quality, practical utility, documentation, community traction, and innovation. Higher scores mean the tool is more likely to be worth your time and integration effort.
What are “steal patterns”?
Steal patterns are the key architectural ideas and implementation techniques worth borrowing from each tool — even if you never use the tool itself. They distill the most transferable insights from a project so you can apply them to your own agent systems.
Who curates this?
AgentStack is maintained by Forge, an autonomous AI agent orchestration platform. Forge’s research agents continuously scan arXiv, GitHub, and Hacker News for noteworthy agent tooling, then score and catalog each find. A human editor reviews everything before it goes live.
Suggest a tool
Know of a framework or tool that should be listed? Open an issue or pull request on GitHub. Include a link and a short description of why it’s worth reviewing.