heimdall-mcp-server

Overview

Heimdall is an MCP server that creates a persistent, growing memory for AI coding assistants. It indexes project documentation, git history, and development insights, allowing LLMs to recall specific solutions and patterns across conversations. The server uses Qdrant for vector similarity search and SQLite for metadata storage, providing tools for storing, retrieving, and managing memories with semantic search capabilities.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• store_memory

  Stores a new piece of information, such as an insight or a solution.

  Try:Remember that we use JWT tokens for authentication in our APITry:Store the solution for the database connection pooling issue we just resolved
• recall_memories

  Performs a semantic search for relevant memories based on a query.

  Try:Find information about our authentication systemTry:Recall any previous solutions for handling concurrent database connections
• session_lessons

  Records a key takeaway from the current session for future use.

  Try:Note that we should always validate user input on both client and server sidesTry:Remember that the bug was caused by an unexpected null value in the response
• memory_status

  Checks the health and statistics of the memory system.

  Try:How many memories do we currently have stored?Try:Check the memory system health and statistics
• delete_memory

  Delete a specific memory by its unique ID.

  Try:Delete memory with ID 12345Try:Remove that outdated authentication method explanation
• delete_memories_by_tags

  Delete all memories that have any of the specified tags.

  Try:Delete all memories tagged with 'temp-analysis'Try:Remove all task-specific memories

Comparable tools

Installation

pip install heimdall-mcp

# Initialize project
heimdall project init

# For Claude Desktop, add to config.json:
{
  "mcpServers": {
    "heimdall": {
      "command": "python",
      "args": ["-m", "heimdall.mcp"]
    }
  }
}

FAQ

What database does Heimdall use?

Heimdall uses Qdrant for vector similarity search and SQLite for metadata storage.

How does Heimdall integrate with git history?

Heimdall indexes entire git history and can optionally install git hooks to automatically update memories on commits.

Compare heimdall-mcp-server with