mem0-bridge

name	mem0-bridge
description	Bridges the mem0 MCP server to OpenClaw's memory system, implementing retrieve-before-plan and store-after-phase patterns for persistent cross-session knowledge.

mem0 Bridge Skill

Status: READY (requires mem0 MCP server running)

Architecture

This skill bridges the mem0 MCP memory server (used by Cursor IDE) with OpenClaw's built-in memory system. It implements the memory contract from docs/ai/memory/MEMORY_CONTRACT.md:

• Retrieve before plan: at session start, query mem0 for relevant project/user context
• Store after phase: after each completed phase, persist key decisions and facts
• Conflict resolution: repo docs (docs/ai/) always win over stored memories

mem0 Integration Pattern

┌──────────────┐     ┌──────────────┐     ┌──────────────┐
│  OpenClaw    │────▶│  mem0 Bridge │────▶│  mem0 MCP    │
│  Agent       │◀────│  Skill       │◀────│  Server      │
└──────────────┘     └──────────────┘     └──────────────┘

Capabilities

• Search memories by semantic query
• Store new observations (atomic facts, ≤120 chars, present tense)
• Deduplicate before storing (search first, write only missing)
• List entities and their observations
• Bridge OpenClaw session context to mem0 graph

Memory Rules (from MEMORY_CONTRACT.md)

• One observation per fact, ≤120 chars, present tense
• Never persist: secrets, tokens, API keys, credentials, personal data
• Persist: stacks, frameworks, naming conventions, key decisions
• Repo docs win over stored memories in conflicts

Configuration

{
  skills: {
    entries: {
      "mem0-bridge": {
        enabled: true,
        env: {
          MEM0_API_URL: "http://localhost:8080",
        },
      },
    },
  },
}

Unblock Steps

1. Ensure mem0 MCP server is running (check Cursor MCP settings)
2. Configure mem0 API endpoint in skill config
3. Run openclaw config set skills.entries.mem0-bridge.enabled true