rlm-claude

Overview

RLM solves Claude Code's context window limitation by providing an MCP server that automatically saves conversation history before compact operations. It implements two memory systems: Insights for key decisions and facts, and Chunks for full conversation segments. The server offers advanced features including semantic search with BM25 + cosine similarity, fuzzy search, multi-project support, smart retention policies, and sub-agent analysis capabilities. RLM hooks into Claude Code's events to provide zero-configuration persistent memory across sessions.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• rlm_remember

  Save decisions, facts, preferences with categories and importance levels

  Try:Remember that the client prefers 500ml bottlesTry:Save the decision to use PostgreSQL for the backendTry:Remember that we should always deploy LOCAL → VPS
• rlm_recall

  Search insights by keyword, category, or importance

  Try:What did we decide about the API architecture?Try:Show me critical insights about deploymentTry:Find facts about authentication systems
• rlm_forget

  Remove an insight from memory

  Try:Forget the temporary note about testing credentialsTry:Remove the old pricing decision from insights
• rlm_status

  Get system overview including insight count and chunk stats

  Try:What's the current status of my memory?Try:Show me memory statistics
• rlm_chunk

  Save conversation segments with typed categorization

  Try:Save this discussion about API redesign as a session chunkTry:Take a snapshot of our current conversationTry:Save this debug information for later reference
• rlm_peek

  Read a chunk in full or partially by line range

  Try:Show me the first 50 lines of chunk 2026-01-18_MyProject_001Try:Read the full conversation chunk from yesterday
• rlm_grep

  Regex search across all chunks with optional fuzzy matching

  Try:Search for 'authentication' across all chunksTry:Find mentions of 'bug #42' with fuzzy matching
• rlm_search

  Hybrid search using BM25 + semantic similarity across chunks and insights

  Try:Search for information about our API architecture decisionsTry:Find content related to deployment strategies
• rlm_list_chunks

  List all chunks with metadata

  Try:Show me all my saved conversation chunksTry:List all session chunks from this week
• rlm_sessions

  Browse sessions by project or domain

  Try:Show me all sessions for MyApp projectTry:List sessions from the backend domain
• rlm_retention_preview

  Preview what would be archived in a dry-run

  Try:What chunks would be archived if I ran retention now?Try:Preview the purge operation
• rlm_retention_run

  Archive old unused chunks and purge ancient ones

  Try:Run the retention process to clean up old chunksTry:Archive unused chunks and purge anything older than 6 months

Comparable tools

Installation

pip install mcp-rlm-server[all]

uv tool install mcp-rlm-server[all] --python 3.12

{
  "mcpServers": {
    "rlm": {
      "command": "python",
      "args": ["-m", "mcp_server.server"]
    }
  }
}

FAQ

How does RLM persist memory between sessions?

RLM hooks into Claude Code's /compact event to automatically save snapshots before context loss. Memory is stored locally in ~/.claude/rlm/ with JSON for insights and chunk files for conversation history.

What's the difference between Insights and Chunks?

Insights store key decisions, facts and preferences with categories and importance levels. Chunks store full conversation segments typed as snapshot, session or debug. Insights are loaded by importance, chunks are searched and retrieved on demand.

Compare rlm-claude with