Log inSubmit server
Mnemosyne OS logo

Mnemosyne OS

Local MCP memory server for searching, asking, and storing agent context across governed vaults.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install Mnemosyne OS

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "mnemosyne-os": {
      "command": "npx",
      "args": [
        "-y",
        "@mnemosyne_os/mcp"
      ]
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Install not yet sandbox-tested— the server itself is confirmed live (we called its tools/list endpoint). The stdio install command is a separate, ongoing check.

Overview

The Mnemosyne OS MCP server connects AI clients to locally managed memory vaults, project context, and selected workspace tools. Vault protection modes, isolated sandboxes, and governance instructions define how agents may access memory. Reach for it when an agent needs persistent, machine-local context across sessions rather than one-off conversation history.

Use cases

  • Search prior decisions and session summaries before continuing work
  • Store architecture notes and debugging findings for later retrieval
  • Check other agents for shared working-tree changes before committing
  • Track project position and next steps across coding sessions
  • Manage local tasks and calendar appointments from an AI client

Key features

  • Vault-scoped memory search and ingestion
  • Grounded question answering with source memories
  • Governance-aware vault protection and spine classification
  • Coding-agent activity and collision inspection
  • To-do and calendar management
  • Resonance, cockpit, and Pheme controls

Tools25

What Mnemosyne OS lets your agent do. Expand a tool for its parameters.

Verified live~7.8k tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked 2d ago

Recent health check succeeded.

8/8 checks healthy over the last 47d

What it can do

From the 25 tools this server returned when we ran it; 25 declare their own safety hints.

Runs commands · 1Can delete or modify data · 4Handles credentials · 2Makes network requests · 1Read-only tools · 20
Network at startup3 connections
Files written at startupNone
Install scriptsNone
npm provenanceNot published
Package age101 days
Maintainers1

Observed 2026-10-19 in an isolated sandbox. A description of reach, not a security audit.

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 1d ago via OSV.dev · @mnemosyne_os/mcp (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars20Stargazers on the source repository.
npm downloads1.1kPackage downloads in the last 30 days.
Last commit3d agoMost recent push to the default branch.
Tools exposed25Callable tools this server registers over MCP.
Directory activity1 viewsConfig copies, upvotes, and views on AllMCPs.

How Mnemosyne OS works

What Mnemosyne OS MCP server does

The Mnemosyne OS MCP server gives AI clients access to persistent memory stored on the user’s machine. Its primary data boundary is the vault: a configured memory store with a token, display name, and memory count. Agents can list available vaults, search them, ask questions grounded in their contents, and write new memories such as decisions, architecture notes, debugging findings, or session summaries.

This fits the 🧠 Knowledge & Memory category because the server is built around durable context across sessions. It also exposes project and workspace information, including recent Git commits, resonance positions, memory classifications, Dream State connections, and activity from other coding-agent sessions. Separate tools can update the user’s To-do backlog, calendar, cockpit status, and Pheme radar.

How it works

The Mnemosyne OS MCP server exposes named MCP tools to connected clients. mnemosyne_vault_list should be used when the target vault is unknown; its returned token is passed to the vault-oriented tools. The server includes governance instructions covering vault protection, the NORMAL and MAXIMUM modes, mixableWith, isolated sandbox vaults, the spine model, and work involving another person’s memory.

For retrieval, mnemosyne_memory_query returns matching memories word for word. Its ranking combines vector similarity with a local BM25 channel and weights results by spine type, making it suitable when the original text must be quoted or used as documentation. mnemosyne_memory_ask runs deeper retrieval, lexical fusion, and re-ranking, then returns a prose answer with its sources. The prose is a model rewording, so source memories should be checked before relying on it.

Writing is explicit: mnemosyne_memory_ingest stores a memory permanently and indexes it for later retrieval. Resonance tools record project continuation points, while agent-collision and file tools inspect shared working-tree activity using transcript files on disk. The Git log tool reads the repository configured by the application and cannot be redirected to another checkout.

Setup and configuration

It identifies Claude Code, Cursor, and Claude Desktop as clients that can connect, and the release notes also mention Codex. The server’s usable vaults are configured on the Mnemosyne OS side; vaults not configured there are flagged and refused until the user adds them.

Some capabilities depend on the application state. Cockpit updates require the app window to be open. Pheme tools require the app to be running, although Pheme itself may be closed. To-do and agenda tools can work with the app closed on a development install through its headless daemon; an npm install requires the app to be running for those operations.

Tools and capabilities

  • Search vaults for exact source memories or ask grounded questions across multiple memories.
  • Ingest durable memories and record resonance phases, notes, and next steps.
  • List vaults, spine assignments, Dream Bridges, and recent configured-repository commits.
  • Inspect other agent sessions, recent file activity, and possible shared working-tree collisions.
  • Add, read, update, and remove To-do items and calendar appointments, subject to each tool’s ID and deletion rules.
  • Update cockpit status or manage Pheme radar sources and scan results.

Limitations and notes

The Mnemosyne OS MCP server does not treat retrieval scores as confidence values. A result and a miss may have similar numbers, so the returned text and source list require review. mnemosyne_memory_ask can take about 30 seconds, and its prose should not be quoted as if it were original memory.

Calendar removals are permanent, and repeating appointments are removed as whole series. To-do removals archive tasks by default, while permanent deletion requires an explicit option. Agent-session inspection depends on transcript files: sessions whose transcripts are elsewhere may not appear, and silence cannot distinguish an idle agent from a crashed one.

The Git tool needs both the application-configured repository and the monorepo:read scope. The Dream State engine must be enabled in Settings and run during idle time before it can produce bridges. The Mnemosyne OS MCP server therefore suits users who want local, governed memory and workspace awareness, but its available data is limited to configured vaults, repositories, files, and application features.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Mnemosyne OS

Mnemosyne OS is an MCP server that connects AI clients to locally managed memory vaults and selected workspace data. Its main tools list vaults, search or ask over memories, ingest durable context, inspect agent activity, read Git history, and manage tasks or calendar entries. Access follows configured vaults and governance rules.

Use the documented npm command `npx -y @mnemosyne_os/mcp` to connect the server. The README identifies Claude Code, Cursor, and Claude Desktop as supported connection targets, and its release notes also mention Codex. Vault availability is configured through Mnemosyne OS.

Yes, Cursor is named in the README as a client that can connect to Mnemosyne OS through MCP. The documented connection uses `npx -y @mnemosyne_os/mcp`; the server then exposes configured vault and workspace tools to the client.

Use `mnemosyne_memory_query` when you need matching memories returned word for word, or `mnemosyne_memory_ask` when you need a deeper answer grounded in multiple memories. Call `mnemosyne_vault_list` first if you do not know the vault token.

Yes, it provides tools for reading and changing To-do lists and calendar appointments. Read tasks or appointments first to obtain their IDs before updating or removing them. On an npm install, the app must be running for these operations; a development install can use the headless daemon with the app closed.

Related MCP servers

All in Knowledge & Memory All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewMnemosyne OS AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/mnemosyne-os?style=directory)](https://allmcps.com/mcp/mnemosyne-os)
HTML Embed
<a href="https://allmcps.com/mcp/mnemosyne-os"><img src="https://allmcps.com/api/badge/mnemosyne-os?style=directory" alt="Mnemosyne OS on AllMCPs" /></a>