I built a local project brain because coding agents kept forgetting the same repository
I wanted a simpler handoff between AI coding sessions:
- Point a local tool at a repository
- Keep architectural decisions, evidence, and unfinished work connected
- Know when remembered context has become stale
- Generate a focused context pack for the next task instead of pasting the whole project into chat
That became Rta-Smriti Brain. It maps files, symbols, imports, memories, checkpoints, and provenance, then prepares bounded context for Codex, Claude Code, Cursor, or another coding agent.
The v0.4.0-alpha release runs locally on Windows, macOS, and Linux. It uses local SQLite storage and has no hosted account or telemetry.
It is still an alpha, so I am especially interested in installation failures, confusing workflows, and cases where the retrieved context is not useful. Those reports are more valuable right now than a generic “looks good.”
Product Hunt:
https://www.producthunt.com/products/rta-smriti-brain?launch=rta-smriti-brain
GitHub:
Replies