memory-map

Installs and configures persistent cross-session memory for Claude Code via an MCP server.

1|Updated Mar 18, 2026
One-click install
npx skills add https://github.com/MARUCIE/openclaw-foundry --skill memory-map-marucie
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: memory-map
Source: https://github.com/MARUCIE/openclaw-foundry/tree/main/web/public/packs/spellbook-ai-app-engineer/skills/memory-map
Command: npx skills add https://github.com/MARUCIE/openclaw-foundry --skill memory-map-marucie

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Claude Code starts every session with no memory of prior work, forcing users to re-explain project context, preferences, and decisions. This Skill installs and configures memory_map, an MCP server that gives Claude persistent key-value memory and rolling conversation history across sessions and projects. ## Core Features & Use Cases - Persistent Key-Value Memory: Store project facts like stack, gotchas, and current work in .mcp_memory.json, loaded automatically at session start via CLAUDE.md instructions. - Rolling Conversation History: Lifecycle hooks (UserPromptSubmit, PreCompact, Stop) automatically save history chunks to .mcp_history.json, with a /mem_save manual checkpoint command. - Cross-Project Recall: Search and load memory across all projects on the machine, plus global memory for user-level preferences. - Use Case: After a week away from a project, Claude opens a new session, calls load_memory and load_history, and immediately knows the stack, current blockers, and where the last conversation left off. ## Quick Start Install and register the memory_map MCP server for Claude Code, then set up the session-start memory loading instructions in my project.

Frequently Asked Questions about memory-map

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I give Claude Code persistent memory across sessions?▼

Install the memory_map MCP server, register it with `claude mcp add -s user`, and add lifecycle hooks to settings.json. Then copy the provided CLAUDE.md into your project so Claude calls load_memory and load_history at every session start.

How do I set up an MCP server for conversation history in Claude Code?▼

Clone the memory_map repo, create a Python virtualenv, install requirements, and register server.py as an MCP server. Add UserPromptSubmit, PreCompact, and Stop hooks pointing to history_hook.py so history saves automatically.

Should I register an MCP server with user or project scope?▼

Use `-s user` scope for memory_map because it stores files at local absolute paths that differ per machine. Project scope commits the config to the repo and breaks for teammates with different paths.

Why does Claude start sessions with no memory after installing an MCP server?▼

MCP registration only makes tools available; it does not call them. Claude auto-loads memory only if the project's CLAUDE.md contains session-setup instructions telling it to call load_memory and load_history first.

Is it safe to enable external summarization for conversation history?▼

History is stored locally by default with simple truncation. Enabling MCP_HISTORY_EXTERNAL_SUMMARIZE sends up to 4,000 characters of dialogue, including code and file contents, to OpenAI, so only enable it after checking your organization's data policy.