#747 · Primary category: AI Agents & Automation

vestige

ai-memory claude cognitive-science cursor embeddings fsrs local-first long-term-memory mcp mcp-server neuroscience onnx rust security spaced-repetition sqlite sveltekit threejs vscode webgpu

Vestige enhances agents by deterministic root-cause retrieval that reaches backward through time to find the quiet change, decision, or service that caused today’s failure, not the lookalike.

Project last updated:08/30/26

GitHub Stars

614

Forks

62

Contributors

15

License

AGPL-3.0

Why we included this project

Agents that run for a while tend to hit the same wall: they re-learn lessons you already taught them and re-derive fixes that are already written down. Vestige is a memory layer that addresses this, running as a local MCP server so Claude Code, Cursor, Codex, and other MCP-capable agents can write memories during a session and pull them back later, with everything stored in a SQLite file on your own machine and no cloud dependency. The retrieval logic is what makes it more than a note store. It merges redundant memories, supersedes contradicted ones, and lets unused entries fade, while also tracing a current failure back to the earlier decision that actually caused it, even when the cause and the symptom share no vocabulary. After a one-time model download it runs fully offline as a single 25MB Rust binary, which makes it a practical choice for teams that cannot send agent data to a cloud service.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category