One shared memory for every agent in a project, over MCP

Description and use

A shared memory layer for all project agents, implemented locally with Mem0 and Qdrant, eliminates external API keys and provides a unified context.

Problem

Multiple AI coding agents in a project need a consistent, shared memory layer without external API keys or complex setup.

Outcome

All agents share a single memory layer, reducing duplication, eliminating API key management, and enabling consistent context across the project.

Workflow

• 1. Install Mem0 and Qdrant locally.
• 2. Clone the agent-memory repository from GitHub.
• 3. Run the provided scripts to generate configuration files for five agents.
• 4. Seed the memory layer with shared rules once.
• 5. Launch agents; they automatically connect to the shared memory via MCP.
• 6. Agents read and write to the shared memory, ensuring consistent context across the project.

Evidence

Universal MCP memory layer for AI coding agents (OpenCode, Claude Code, Cursor, Kilo Code). Self-hosted with Mem0 + Qdrant. Seed rules once, all agents remember.

Source