Setting up Dakera MCP for Claude Code / Cursor / Windsurf

[ferhimedamine]

Quick Setup Guide

Here's how I set up Dakera MCP to give my AI coding assistants persistent memory across sessions.

1. Start the memory server

git clone https://github.com/dakera-ai/dakera-deploy.git
cd dakera-deploy
docker compose up -d

Server starts on localhost:3300 with embedded storage (no external DB needed).

2. Install the MCP server

cargo install dakera-mcp

3. Configure your client

Claude Code (~/.claude/settings.json):

{
  "mcpServers": {
    "dakera": {
      "command": "dakera-mcp",
      "env": {
        "DAKERA_API_URL": "http://localhost:3300",
        "DAKERA_API_KEY": "your-api-key"
      }
    }
  }
}

Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "dakera": {
      "command": "dakera-mcp",
      "env": {
        "DAKERA_API_URL": "http://localhost:3300",
        "DAKERA_API_KEY": "your-api-key"
      }
    }
  }
}

What you get

68+ MCP tools for memory management:

• dakera_store / dakera_recall — persist and retrieve memories
• dakera_hybrid_search — BM25 + vector search in one call
• dakera_knowledge_graph — entity extraction and relationship mapping
• dakera_session_start / dakera_session_end — session lifecycle
• dakera_consolidate — merge related memories
• dakera_knowledge_deduplicate — clean up similar memories

Architecture

The memory server runs a decay-weighted importance model — memories naturally fade unless they're accessed frequently. This means your agent's memory stays relevant without manual cleanup.

Full deploy options (including Kubernetes and HA): https://github.com/dakera-ai/dakera-deploy

---

Happy to answer questions about the setup or architecture!