MegaMemory

Overview

MegaMemory allows coding agents to build and query a graph of concepts, architecture, and decisions across sessions. Using LLM-generated concepts rather than AST parsing, it creates a knowledge graph that persists in SQLite. The server provides tools for understanding project context, creating/updating concepts, managing relationships, and resolving merge conflicts between different versions of the knowledge graph. It includes a web explorer for visualization and supports multiple MCP clients.

Try asking AI

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

• understand

  Semantic search over the knowledge graph. Returns matched concepts with children, edges, and parent context.

  Try:Find all modules related to authenticationTry:Search for concepts about database indexing patterns
• get_concept

  Look up a concept by its exact ID. Returns full context including children, edges, incoming edges, and parent.

  Try:Get concept with ID user-auth-moduleTry:Retrieve concept details for payment-processing-2023
• create_concept

  Add a new concept with optional edges and file references.

  Try:Create a new concept for the user authentication featureTry:Add a concept about the new caching implementation
• update_concept

  Update fields on an existing concept. Regenerates embeddings automatically.

  Try:Update the concept about the payment processor to include new error handlingTry:Modify the user session concept to add references to auth files
• link

  Create a typed relationship between two concepts.

  Try:Link the auth concept to the user database concept with depends_on relationshipTry:Connect the payment module to the notification system with calls relationship
• remove_concept

  Soft-delete a concept with a reason. History preserved.

  Try:Remove the deprecated concept old-login-system because it's been replacedTry:Delete the concept about abandoned feature X
• list_roots

  List all top-level concepts with direct children.

  Try:Show all top-level conceptsTry:List the main architecture components
• list_conflicts

  List unresolved merge conflicts grouped by merge group.

  Try:Show all merge conflictsTry:List conflicts in the knowledge graph
• resolve_conflict

  Resolve a merge conflict by providing verified, correct content based on the current codebase.

  Try:Resolve conflict for user management module with correct implementationTry:Fix the merge conflict for payment processing with current code

Comparable tools

Installation

npm install -g megamemory

{
  "mcpServers": {
    "megamemory": {
      "command": "megamemory",
      "args": []
    }
  }
}

megamemory install --target <client-name>

FAQ

How does MegaMemory store knowledge?

It uses SQLite with WAL mode, storing concepts as nodes with relationships as edges. Each project has its own database at .megamemory/knowledge.db.

Does it require API keys or network calls?

No, embeddings are processed in-memory using the all-MiniLM-L6-v2 model that downloads once (23MB). No API keys needed.

Compare MegaMemory with