Hermes Atlas
Memory & knowledge · works with Hermes Agent

MemoryBuddy

Trainspotting31/memory-buddy

Shared memory server on Cloudflare Workers that Hermes and other MCP clients read and write

In short

MemoryBuddy is a shared memory layer for AI agents, deployed on Cloudflare Workers and exposed over MCP. Hermes Agent, Trae, Cursor and Claude Desktop can connect to the same server and read and write one common memory.

What MemoryBuddy does

MemoryBuddy runs as a single Cloudflare Worker with three storage tiers. A Durable Object holds the current conversation, a D1 database stores structured facts such as names and preferences, and a Vectorize index holds embeddings for search by meaning. An LLM extracts the facts worth keeping, and long conversations are summarized so they stay compact.

The Worker exposes an MCP server at /mcp with five tools over Streamable HTTP, an HTTP chat API at /chat with SSE streaming and automatic fact extraction, and a REST API at /memory/:userId. A DELETE call removes everything stored for a user. Hermes Agent connects with a hermes mcp add command pointing at the Worker's /mcp URL, and the README says the whole setup runs on Cloudflare's free tier.

Key features

  • MCP server with five tools over Streamable HTTP
  • Semantic recall of memories through Cloudflare Vectorize
  • Automatic fact extraction and conversation summarization
  • Three-tier memory: Durable Object, D1 facts and Vectorize embeddings
  • REST API plus a single DELETE call that wipes a user's memory
  • Runs on Cloudflare's free tier according to the README

When to use it

  • Sharing preferences and context between Hermes Agent and an editor such as Cursor or Trae
  • Keeping long-term facts outside any one agent so switching tools does not lose context
  • Running a self-hosted memory backend on a free Cloudflare account

Who it is for: Developers who use several MCP-capable AI tools and want one self-hosted memory they control.

How it fits with Hermes Agent

Supports Hermes Agent as one MCP client among several; Hermes connects with hermes mcp add memory-buddy and the URL of your deployed Worker's /mcp endpoint.

How to install MemoryBuddy

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

git clone https://github.com/Trainspotting31/memory-buddy.git
cd memory-buddy
npm install
npx wrangler deploy

Requirements: A Cloudflare account (free tier is fine) and Node.js 18+

FAQ

What is MemoryBuddy?

MemoryBuddy is a self-hosted shared memory service for AI agents that runs on Cloudflare Workers. Any MCP-compatible tool can read and write the same long-term memory through it.

Does MemoryBuddy work with Hermes Agent?

Yes. The README lists Hermes Agent as a supported client and connects it with the hermes mcp add command pointed at your Worker's /mcp URL.

How do I install MemoryBuddy?

Clone the repository and run npm install. Then create the D1 database and Vectorize index with wrangler, apply schema.sql, copy the database_id into wrangler.toml, and run npx wrangler deploy.

Similar memory for Hermes Agent

All memory

Related guides: SOUL.md for Hermes Agent: what it is and how to write one · Run multiple Hermes agents with profiles