Building a persistent memory layer for AI agents using fast‑mcp

[Pepan]

Hi everyone! 👋 I wanted to share something I’ve been working on over the past few weeks. As more developers build agents on top of large language models, one recurring pain point keeps cropping up: context disappears between sessions. Agents start each chat or code session with no recollection of what happened before.

To solve this, I built Memory Notes – a simple MCP‑compliant server that gives your agents a persistent, semantic memory. It’s built on top of fast‑mcp, so we can host and extend it easily.

####Why Memory Notes?

• **Memories are shareable containers (think of them as projects or workspaces). You can invite other users or agents to collaborate in the same memory.
• Notes live inside a memory. Agents (or humans) can write, and search these notes via the Model Context Protocol. Under the hood, notes are semantically indexed, so you can retrieve them by meaning instead of exact words.
• It’s built with fast‑mcp, which made implementing the server straightforward and standards‑compliant. The gem handles the protocol plumbing so you can focus on your business logic.

####Try it out

Memory Notes is live as a free public beta. If you want to see an agent with memory in action or integrate it into your own projects, visit llm‑memory.com. There’s also a page for creators and developers that includes quickstart guides and example payloads: llm‑memory.com/creators.

I’d love to hear your feedback or see what you build on top of it. Feel free to leave a comment or open an issue if you have ideas or questions!