About
Agentic Stacks is a neutral directory of the components teams use to build AI agents — models, frameworks, harnesses, memory, retrieval, eval, observability, MCP servers, and more — alongside real, cited stacks that show how those pieces fit together in production.
Today it indexes 179 tools across 16 categories and 31 stacks. Every entry is tagged with where it came from, so you can discover what's working, compare alternatives, and cite what you read.
Why no fabricated metrics?
Most directories pad themselves with invented numbers — made-up eval scores, made-up "cost per request," made-up adoption stats. We don't.
A directory is only as useful as it is trustworthy. We index real component choices and architectural rationale. When real teams submit real metrics, those numbers appear on the relevant stack — clearly attributed to their source.
Provenance tiers
- public_docsSourced from vendor docs and official READMEs.
- researchedReal stack reconstructed from public talks, blog posts, OSS code; cited.
- referenceSynthesized reference architecture; not attributed to any real company.
- communitySubmitted by a real team using a real production stack.
Open data
The full corpus is available as JSON and as YAML files in the source repo. Build on it — embed widgets, comparison pages, agent tools that consume the data, whatever you want.
Community
Agentic Stacks is built to be community-maintained: the corpus lives as YAML and will be open on GitHub, so anyone can add tools, fix descriptions, or propose new categories through a pull request. Community submissions open soon.