Your agent's memory shouldn't die when you switch tools. Brain Memory is one shared brain for Claude Code, Codex CLI, OpenCode, Copilot, Antigravity + the Claude; ChatGPT apps via MCP — change model, vendor, device or agent and pick up where you left off. Modeled on neuroscience: memories decay on a forgetting curve, strengthen with recall, link associatively, consolidate during sleep. Plain Markdown files, no database, deterministic recall. MIT plugin, free forever; optional cloud sync $1/mo.
No reviews yetBe the first to leave a review for Brain Memory
Maker
📌
Hey Product Hunt 👋
I'm the solo founder of Brain Memory.
I built it because my agent forgot everything the moment I switched tools — new model or new IDE, and I'm re-explaining my whole project from scratch. So I gave it one memory that travels with me: the same brain across every model, tool, and device. No export, no migration, no starting over.
It works like human memory, not a database. Memories organize themselves — grouping by topic and linking to related ones as you go deeper. Unused ones fade on the Ebbinghaus forgetting curve; the ones you keep recalling grow stronger and reorganize over time.
I benchmarked it against real baselines: on a 1,000-distractor haystack brain passes 100% where a vector store scores 0% — at the leanest tokens-per-success of any passing arm. Small sample (n=3), honest nulls included, full tables + raw JSON at brainmemory.ai/docs/benchmarks/results.
The file system is the database — plain Markdown + YAML. Open your agent's memory in any text editor, grep it, or version it with git.
MIT-licensed, fully open source, zero npm dependencies, and free forever. If you want your brain synced across devices there's an optional cloud tier at $1/mo — the local plugin never needs it.
Would love your feedback — I hope it makes your (and your agents') work a little less forgetful.