Problems and solutions
ShellBrain stores each problem with the solutions that worked and the tactics that failed.
ShellBrain
Long-term memory for AI agents.
ShellBrain gives AI agents memory, so they can store and reuse what they learn over time. Works with Codex, Claude Code, and Cursor.
$
curl -L shellbrain.ai/install | bash
Recall
Recall in one command.
How it works
Two inner agents.
Working Agent
Visible
Your agent runs shellbrain recall, receives a compact brief, and keeps working.
Inside ShellBrain
Knowledge Builder
Inner Agent
After each session, it turns useful work into durable memories backed by evidence.
Recall Agent
Inner Agent
It searches those memories and returns only the context the current task needs.
What it remembers
Grounded in evidence.
Facts and preferences
It records durable facts about the repo and your preferences. Changes create a visible revision history.
Evidence
Every memory links to the session evidence that supports it. Ambiguous memories are skipped.
Current state
Memories can be active, stale, superseded, or wrong. Recall weighs that state before returning context.
Your data
Your memory stays yours.
ShellBrain runs locally with no cloud retrieval or outbound telemetry.
Store data on your machine or in your own PostgreSQL database.