
One memory shared by Claude Code, Codex, Cursor, Copilot CLI, OpenClaw and 20 more coding agents, bu
One memory shared by Claude Code, Codex, Cursor, Copilot CLI, OpenClaw and 20 more coding agents, bu
One memory shared by Claude Code, Codex, Cursor, Copilot CLI, OpenClaw and 20 more coding agents, built from the session history already on disk. A fix found in one agent comes back in any of them, including months of sessions from before you installed it. No LLM, no embeddings, one local Go binary.
wire the six stores that landed read-only, surface by surface
parser: JetBrains AI Assistant and Junie chats (workspace XML + aia-task-history event logs)
copilot-chat: VS Code 1.137 writes chats to extension transcripts + session-store.db, so the source finds 0 files
git blame says who changed a line; nothing says why the agent did it that way
Tell people about the secrets already sitting in their agent transcripts
asked-twice is blind to instructions: 9 of 955 Claude sessions carry an asked hash
the docs site still lists three harness packages; there are four
only search suggests a flag when you mistype one; the other 22 sites just refuse
doctor reports a stale Kimi plugin but says nothing about a stale Grok one
deja index has no --quiet, so a scheduled run either prints a line or hides its errors