Recall FAQs

Question 1

How does Recall automatically capture and store information?

Accepted Answer

Recall operates in the background, using a 'Stop Hook' that fires at the end of each Claude Code session. This triggers an 'Extraction pipeline' which leverages Claude Haiku (or Ollama fallback) to parse and summarize the conversation. The extracted data is then stored in a SQLite database and markdown files.

Question 2

What types of search capabilities does Recall offer for finding past memories?

Accepted Answer

Recall provides versatile search options: full-text keyword search (via FTS5), semantic search using Ollama embeddings for conceptual relevance, and a powerful hybrid search (default) that combines both methods using Reciprocal Rank Fusion for highly accurate results.

Question 3

How does Recall provide context at the start of new Claude Code sessions?

Accepted Answer

Recall implements 'Tiered session-start context'. This includes L0 identity (your role, projects, preferences from `identity.md`) and L1 important records (top 12 records ranked by an importance score), which are automatically loaded to give Claude Code immediate relevant background.

Question 4

What is Recall and what problem does it solve for Claude Code users?

Accepted Answer

Recall is a tool that provides persistent memory for Claude Code. It solves the problem of Claude Code forgetting context between sessions by automatically extracting, indexing, and recalling critical information, eliminating the need to repeat yourself and ensuring continuity across your AI coding workflow.

Question 5

Can Recall integrate with AI agents or other parts of my AI environment?

Accepted Answer

Yes, Recall features seamless MCP (Modular Context Protocol) integration, allowing Claude Code and any spawned AI agents to access your persistent memory through dedicated tools like `memory_search`, `memory_hybrid_search`, and `context_for_agent`, enhancing their understanding and decision-making.

Recall

Recall

주요 기능

사용 사례

주요 기능

사용 사례