heimdall-mcp-server
If you are the rightful owner of heimdall-mcp-server and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.
Heimdall MCP Server is an AI coding assistant's long-term memory solution, providing persistent, context-rich memory for LLMs.
Heimdall MCP Server addresses the challenge of short-lived memory in AI coding assistants by offering a persistent, context-rich memory system. It integrates with your codebase, learning from documentation, session insights, and development history to provide a comprehensive memory that spans across conversations. By indexing the entire git history of a project, Heimdall understands not only what changes were made but also who made them and when, providing a deep contextual understanding. Each project is given its own isolated memory space, ensuring that context does not leak between projects. Built on the Model Context Protocol (MCP), Heimdall offers a standardized and efficient way for LLMs to access this memory, enhancing their ability to recall specific solutions and architectural patterns.
Features
- Context-Rich Memory: Learns from documentation and development history for comprehensive recall.
- Git-Aware Context: Indexes entire git history for deep contextual understanding.
- Isolated & Organized: Each project has its own memory space to prevent context leakage.
- Efficient Integration: Uses MCP for standardized, low-overhead memory access.
- Real-time Monitoring: Automatic file monitoring and git hooks for seamless updates.
Tools
store_memory
Stores a new piece of information, such as an insight or a solution.
recall_memories
Performs a semantic search for relevant memories based on a query.
session_lessons
Records a key takeaway from the current session for future use.
memory_status
Checks the health and statistics of the memory system.
delete_memory
Delete a specific memory by its unique ID.
delete_memories_by_tags
Delete all memories that have any of the specified tags.