System · MODEL CONTEXT PROTOCOL
tomschell mcp long term memory
A long-term memory storage system for LLMs using the Model Context Protocol (MCP) standard. This system helps LLMs remember the context of work done over the entire history of a project, even across multiple sessions. It uses semantic search with embeddings to provide relevant context from past interactions and development decisions.
WHAT IT CONNECTS
What tomschell mcp long term memory gives an AI Agent
- Project-based memory organization
- Semantic search using Ollama embeddings (nomic-embed-text model, 768 dimensions)
- Multiple memory types:
- Conversations: Dialog context and important discussions
- Code: Implementation details and changes
- Decisions: Key architectural and design choices
- References: Links to external resources and documentation
- Rich metadata storage including:
- Implementation status
- Key decisions
- Files created/modified
- Code changes
- Dependencies added
- Tagging system for memory organization
- Relationship tracking between memories
- Node.js (v18 or later)
- Ollama running locally (for embeddings)
- Must have the `nomic-embed-text` model installed
- SQLite3
- Clone the repository
- Install dependencies:
- Build the project:
- Create a `.env` file with required configuration:
- Start the server in development mode:
- Compile TypeScript
- Copy schema files
- Start the server with auto-reload
- The server connects via stdio for Cursor compatibility
- `projects`: Project information and metadata
- `memories`: Memory entries storing various types of development context
- `embeddings`: Vector embeddings (768d) for semantic search capabilities
- `tags`: Memory organization tags
- `memory_tags`: Many-to-many relationships between memories and tags
- `memory_relationships`: Directed relationships between memory entries
- `store-dev-memory`: Create new development memories with:
- Content
- Type (conversation/claude-code/decision/reference)
- Tags
- Code changes
- Files created/modified
- Key decisions
- Implementation status
- `list-dev-memories`: List existing memories with optional tag filtering
- `get-dev-memory`: Retrieve specific memory by ID
- `search`: Semantic search across memories using embeddings
- Kill any existing server instances
- Rebuild the TypeScript code
- Copy the schema.sql to the dist directory
- Start the server in development mode
- `@modelcontextprotocol/sdk@^1.7.0`: MCP protocol implementation
- `better-sqlite3@^9.4.3`: SQLite database interface
- `node-fetch@^3.3.2`: HTTP client for Ollama API
- `zod@^3.22.4`: Runtime type checking and validation
- Write clear commit messages
- Add appropriate documentation
- Follow the existing code style
- Add/update tests as needed
SECURITY
An MCP listing is not a security review
An MCP server may receive model context, credentials, local files, or permission to call external systems. Review its code, requested environment variables, network behavior, package provenance, and maintenance status before connecting it to an agent.