Labsco
MCP SERVER

MemGPT MCP Server

by Vic563 ↗

Give a conversation a memory that outlives the session, and change which provider or model is answering without starting over.

Session & Transcript Recall
Summary
One thread, many models, and a history that stays put.

Chat history lives in the server rather than in any one client session, so the same conversation can carry across restarts and across providers. Useful when you want a cheap model for the routine turns and a stronger one for the hard question, without losing what came before.

What it is

A TypeScript MCP server that keeps conversation history on your machine and relays chat to whichever provider you point it at — OpenAI, Anthropic, OpenRouter or a local Ollama model.

What you get
  • Stored conversation history you can read back in chronological order with timestamps, either the last N entries or all of them
  • A clean slate on demand — history can be wiped when a thread has moved on to something unrelated
  • A provider switch that persists: the same conversation continues against OpenAI, Anthropic, OpenRouter or Ollama
  • A model switch inside the chosen provider, so a thread can move between fast models like claude-3-haiku and heavier ones like claude-3-opus
  • OpenRouter addressed in `provider/model` form, and Ollama pointed at whatever you have pulled locally
Requirements

Node with npm — clone, npm install, npm run build, then point your client at the built index.js. Keys go in the client config for whichever hosted providers you use: OpenAI, Anthropic, OpenRouter. Local Ollama models need no key.