MemGPT MCP Server
A server that provides a memory system for LLMs, enabling persistent conversations with various providers like OpenAI, Anthropic, and OpenRouter.
MemGPT MCP Server
A TypeScript-based MCP server that implements a memory system for LLMs. It provides tools for chatting with different LLM providers while maintaining conversation history.
Features
Tools
-
chat- Send a message to the current LLM provider- Takes a message parameter
- Supports multiple providers (OpenAI, Anthropic, OpenRouter, Ollama)
-
get_memory- Retrieve conversation history- Optional
limitparameter to specify number of memories to retrieve - Pass
limit: nullfor unlimited memory retrieval - Returns memories in chronological order with timestamps
- Optional
-
clear_memory- Clear conversation history- Removes all stored memories
-
use_provider- Switch between different LLM providers- Supports OpenAI, Anthropic, OpenRouter, and Ollama
- Persists provider selection
-
use_model- Switch to a different model for the current provider- Supports provider-specific models:
- Anthropic Claude Models:
- Claude 3 Series:
claude-3-haiku: Fastest response times, ideal for tasks like customer support and content moderationclaude-3-sonnet: Balanced performance for general-purpose useclaude-3-opus: Advanced model for complex reasoning and high-performance tasks
- Claude 3.5 Series:
claude-3.5-haiku: Enhanced speed and cost-effectivenessclaude-3.5-sonnet: Superior performance with computer interaction capabilities
- Claude 3 Series:
- OpenAI: 'gpt-4o', 'gpt-4o-mini', 'gpt-4-turbo'
- OpenRouter: Any model in 'provider/model' format (e.g., 'openai/gpt-4', 'anthropic/claude-2')
- Ollama: Any locally available model (e.g., 'llama2', 'codellama')
- Anthropic Claude Models:
- Persists model selection
- Supports provider-specific models:
Development
Install dependencies:
npm install
Build the server:
npm run build
For development with auto-rebuild:
npm run watch
Installation
To use with Claude Desktop, add the server config:
On MacOS: ~/Library/Application Support/Claude/claude_desktop_config.json
On Windows: %APPDATA%/Claude/claude_desktop_config.json
{
"mcpServers": {
"letta-memgpt": {
"command": "/path/to/memgpt-server/build/index.js",
"env": {
"OPENAI_API_KEY": "your-openai-key",
"ANTHROPIC_API_KEY": "your-anthropic-key",
"OPENROUTER_API_KEY": "your-openrouter-key"
}
}
}
}
Environment Variables
OPENAI_API_KEY- Your OpenAI API keyANTHROPIC_API_KEY- Your Anthropic API keyOPENROUTER_API_KEY- Your OpenRouter API key
Debugging
Since MCP servers communicate over stdio, debugging can be challenging. We recommend using the MCP Inspector:
npm run inspector
The Inspector will provide a URL to access debugging tools in your browser.
Recent Updates
Claude 3 and 3.5 Series Support (March 2024)
- Added support for latest Claude models:
- Claude 3 Series (Haiku, Sonnet, Opus)
- Claude 3.5 Series (Haiku, Sonnet)
Unlimited Memory Retrieval
- Added support for retrieving unlimited conversation history
- Use
{ "limit": null }with theget_memorytool to retrieve all stored memories - Use
{ "limit": n }to retrieve the n most recent memories - Default limit is 10 if not specified
相关服务器
Scout Monitoring MCP
赞助Put performance and error data directly in the hands of your AI assistant.
Alpha Vantage MCP Server
赞助Access financial market data: realtime & historical stock, ETF, options, forex, crypto, commodities, fundamentals, technical indicators, & more
Instant Meshes MCP
A 3D model processing server for automatic retopology, simplification, and quality analysis of OBJ/GLB models.
Markdown Navigation MCP
An MCP server that provides efficient navigation and reading of large markdown files using ctags to reduce context usage
Tiktoken MCP
Count tokens using OpenAI's tiktoken library.
E2B
Run code in secure sandboxes hosted by E2B
Postman MCP Server
Interact with the Postman API via an MCP server. Requires a Postman API key.
Abstract MCP Server
Caches large tool responses to files and returns compact resource links to save LLM context window space.
Geo Location Demo
Retrieves user geolocation information using EdgeOne Pages Functions and exposes it via the Model Context Protocol (MCP).
MCP Orchestrator
Aggregates tools from multiple MCP servers with unified BM25/regex search and deferred loading
Add Tool
Fetches and sets up an MCP server from a specified GitHub repository.
Trade-MCP
A modular trading automation project using the Zerodha Kite Connect API for tool-based and resource-based automation.