Install
openclaw skills install @darinrowe/recallbaseRecover context from local AI conversation history
openclaw skills install @darinrowe/recallbaseRecover the smallest useful slice of local AI history, then answer the user's question in natural language.
This workflow requires the local rb CLI on PATH and a RecallBase store populated from supported local sources. Core retrieval runs locally without login or network access.
rb today --jsonrb search "<specific query>" --jsonrb sources --jsonok: true; for ok: false, use error.code, message, and hint to explain or recover.rb open <conversation-id> --json. Search results also provide the stable reference recallbase:conversation/<id>.Retrieval is complete when the evidence answers the question or the available source coverage proves the relevant history is absent or incomplete.
Use an already-configured local RecallBase MCP instead of shell commands when it is available. Read references/mcp.md for MCP routing. The public RecallBase Docs MCP contains product documentation; personal history tools come only from the local rb mcp server.
rb open ... commands as optional follow-up references, not the answer.The response is complete when every material claim is grounded in retrieved history, uncertainty is explicit, and the user can continue without reading command output.
rb is unavailable, retrieval is empty or incomplete, imports fail, a full re-import is required, or browser capture setup is unhealthy.