v0.5.0 — Auto-recall

Auto-recall

Retrieval used to depend on Claude deciding to call search_notes. A model that doesn't suspect a topic is covered never looks, so your notes stay unread.

memex recall install

Every prompt you type is semantically searched against your notes, and the top 3 titles are injected as context before Claude answers — the same way native memory works. Claude then pulls full notes with get_note when a title looks relevant.

A background daemon keeps the embedding model warm behind a unix socket, so a lookup costs ~30ms instead of the ~1.5s a cold search spends loading the model. It idles out after 2 hours.

• New bin memex-recall: hook client, --serve daemon, --warm prewarm
• memex recall install / uninstall writes the hooks into ~/.claude/settings.json, leaving foreign hooks untouched
• Opt-in. ~200MB resident while warm, up to 3 note titles of context per prompt
• Skips prompts under 12 chars and slash/bang commands

Other

• mcpName now ships in the npm package, unblocking MCP Registry ownership verification