# π§ RMS Memory MCP
**Version:** `1.3.0` (2026-10-13) Β· companion GUI `1.3.0` (unified numbering)
**Persistent, local-first memory for your AI coding agents.**
Stop re-explaining your architecture to Cursor, Zed, and Claude Code and other IDEs every single session.
[](https://opensource.org/licenses/MIT)
[](https://www.rust-lang.org/)
[](https://crates.io/crates/rms-memory-mcp)
[](https://github.com/max-ramas/rms-memory-mcp/releases)
[]()
[](https://modelcontextprotocol.io)
[Features](#-key-features) β’ [Download](https://github.com/max-ramas/rms-memory-mcp/releases/latest) β’ [Install](#-installation) β’ [Quick Start](#-quick-start) β’ [CLI](#-cli-commands) β’ [MCP Tools](#-mcp-tools-exposed) β’ [Architecture](#-architecture-highlights)
---
---
## The Problem
You're developing a single project but switching between different agents β Cursor, Zed, Claude Code, OpenCode, etc. Every one of them loses context of architectural decisions, system requirements, and user preferences the moment you close the tab. You end up re-explaining the same things over and over, or copy-pasting a stale `CLAUDE.md` between tools.
**RMS Memory MCP** bridges this gap: a single, isolated, centralized Markdown vault β perfectly structured for LLM consumption β that any MCP-compatible IDE can read from and write to.
### π₯οΈ Prefer a GUI over raw Markdown?
**[RMS Memory GUI](./GUI-README.md)** turns your vault into a visual workspace β full graph editor, per-project Git & GitHub sync, one-click Doctor repair, and an AI-assisted Wiki generator (bring your own key). One-time desktop app, works on top of everything below. The MCP server stays 100% free and standalone either way.
## β¨ Key Features
| | |
|---|---|
| ποΈ **Global Centralized Vaults** | Project context lives outside your repo β zero `.mcp` file pollution. |
| π **Hybrid Retrieval (LanceDB)** | Embedded Vector Search + Tantivy Full-Text Search for zero-fail context hits. |
| π **Multilingual Semantic Parsing** | `fastembed-rs` + `multilingual-e5-small` by default β native Russian & English understanding. Since 1.3.0 `rms-memory embedding use` switches to another bundled model or a local Ollama model (each keeps its own index), and `rms-memory embedding eval` measures hit@k / MRR on your own vault. |
| π³ **AST Markdown Chunker** | `pulldown-cmark`-based chunking keeps code blocks and lists bound to their parent heading. |
| 𧩠**Semantic Code Memory** | Optional Tree-sitter indexing for Rust, Go, JS/JSX, TS/TSX, Python, C/C++, Java, Ruby, Swift, C#, PHP, Kotlin, SQL, HTML, CSS, and Vue `