Skip to content
We’re going live Sunday, April 26 at 9:00 PM CT.
Model Context Protocol

A local MCP server that gives any agent permanent memory across every Claude Code session.

Read tools for search and recall. Write tools for tags and threads. Audit log on by default. Rate-limited. Bound to 127.0.0.1. Built on the open Model Context Protocol so any MCP-aware agent can call it.

Five-minute install

Install Claude Recall (see /install), then add this block to your Claude Code MCP config:

{
  "mcpServers": {
    "recall": {
      "command": "claude-recall-mcp"
    }
  }
}

Restart Claude Code. The agent now has access to every Claude Code session you have ever indexed. Mid-session, it can search past work, fetch context, or apply tags without you breaking flow.

Tools the agent can call

Read

  • search
    Full-text search across every indexed session and note.
  • get_session
    Fetch a session by id with metadata and ordered messages.
  • list_sessions
    List sessions filtered by project, tag, or date range.
  • list_projects
    List indexed projects.
  • list_tags
    List tags and tag counts.
  • context_for_session
    Return condensed context for a session, ready to inject.
  • thread_get
    Fetch a persistent context thread by id.
  • thread_list
    List active and archived threads.
  • thread_for_session
    List threads that reference a given session.

Write

  • apply_tags
    Add tags to a session. Honors the no-delete invariant.
  • list_sessions_to_tag
    Surface untagged or partially-tagged sessions for review.

Write tools never delete. Tag application is additive. Three-layer durability holds for everything written through MCP, the same as for human writes.

Why MCP, not a wrapper API

MCP is the protocol Claude Code already speaks for tool use. Building Claude Recall as a first-class MCP server means agents call it the same way they call file edits or web fetches. No glue code. No custom adapter. No extra latency.

Audit log and rate limits

Every MCP call is appended to an on-disk audit log under ~/.recall/audit/. You can grep it. The MCP server enforces a per-tool rate limit so a runaway agent loop cannot saturate it. Default limits are conservative; configurable in the daemon settings.

Example agent prompt

Drop something like this into a fresh Claude Code session and the agent will reach into your history without you having to drive:

Find the session where I last touched the auth retry logic.
Pull condensed context for it. Tell me what I changed and what was still broken.

The agent calls search, then context_for_session, and answers from the retrieved context. No human session-id juggling.

Privacy posture

Read this next