Picture published by the creator on GitHub.
Give a coding agent memory that learns instead of just recalling
Made by vectorize-io
View profile on GitHub (opens in a new tab)Hindsight is a memory system for AI agents, built to notice patterns across many conversations rather than only replaying past ones the way a typical chat history or a RAG lookup (search over stored documents) does.
You run it as a local server and connect it to a coding agent or your own app through an SDK, an MCP tool, or an LLM wrapper, and it works with over 25 model providers, including fully local ones through Ollama. The creator reports state-of-the-art results on the LongMemEval benchmark, a public test of how well a system remembers across long conversations, and says results have been independently reproduced by outside researchers.
You could use it to…
- Give a coding agent memory that learns, not just recalls
- Run it fully local with Ollama, no cloud key required
- Connect it to an agent over MCP as a memory backend
Can I use this?
You'll need Docker and an LLM provider (local or cloud) · Setup needed
Worth knowing Benchmark figures beyond the independently reproduced one are self-reported by the project, and memory quality still depends on which underlying model you connect it to. Production use is described as already happening at some companies, which is the creator's claim, not an independent audit. MIT licence. BuildTube has not run or verified this software.
Why it's here
Rising fast on GitHub this week
- 18,389 people starred it on GitHub this week — 44,190 in total.
Numbers from the snapshot taken 1 October 2026; not refreshed since.
Behind it
See the code on GitHubBuildTube has not run or verified this project. Everything above is written from what the creator published.
Made this?