Log inSubmit server
MCP Memory Server logo

MCP Memory Server

by mnemoverse

Hosted cross-tool memory with relevance feedback, recency-based recall, shared rooms, and secret alias listing over MCP.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install MCP Memory Server

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": {
    "mnemoverse-mcp-memory-server": {
      "command": "npx",
      "args": [
        "-y",
        "@mnemoverse/mcp-memory-server@latest"
      ]
    }
  }
}

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

mnemoverse/mcp-memory-server connects MCP clients to the hosted Mnemoverse memory engine, allowing durable context to persist across Claude, ChatGPT, Cursor, and VS Code. It stores and searches memories, re-ranks results from helpfulness feedback, lists recent entries, and exposes memory statistics and association graphs. Shared rooms support collaborative context across agents, while room membership controls whether participants can read or write. Use it when agents need user or project context to follow them across sessions and tools; connect through the hosted OAuth endpoint or run the package locally with an API key.

Use cases

  • Store user preferences that should persist across AI tools
  • Recall project decisions and setup details in later sessions
  • Catch up on new memories added to a shared room
  • Tune memory ranking with helpfulness feedback
  • Inspect learned concept associations and memory quality

Key features

  • Cross-tool persistent memory
  • Relevance and recency-based retrieval
  • Feedback-driven memory ranking
  • Shared rooms with scoped membership
  • Recent-memory cursor pagination
  • Memory graph and vault alias inspection

Tools11

What MCP Memory Server lets your agent do. Expand a tool for its parameters.

Verified live~3.6k 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 Today

Recent health check succeeded.

22/25 checks healthy over the last 88d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 2d ago via OSV.dev · @mnemoverse/mcp-memory-server@latest (npm)

Adoption & maintenance

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

GitHub stars28Stargazers on the source repository.
npm downloads4.3kPackage downloads in the last 30 days.
Last commit1d agoMost recent push to the default branch.
Tools exposed11Callable tools this server registers over MCP.
Directory activity11 viewsConfig copies, upvotes, and views on AllMCPs.

How MCP Memory Server works

What mnemoverse/mcp-memory-server does

mnemoverse/mcp-memory-server exposes the hosted Mnemoverse memory engine through the Model Context Protocol. It gives agents a durable store for preferences, project decisions, facts about people or roles, and lessons learned. Memories remain available across sessions and across connected tools using the same Mnemoverse account.

The basic workflow uses memory_write to save a self-contained statement and memory_read to retrieve relevant memories with a natural-language query. An importance check can reject low-value writes, and the write result indicates whether storage occurred. Agents can send the identifiers of recalled memories to memory_feedback: positive feedback improves their future ranking, while negative feedback lowers their ranking without deleting them.

The service is intended for durable context rather than transient conversation. Passwords, API keys, payment information, MFA codes, government identifiers, and health records should not be stored. A corrected or outdated memory is superseded by writing a newer memory rather than deleting the original.

How it works

The hosted MCP endpoint is https://mcp.mnemoverse.com/mcp. Clients that support remote MCP connections can authenticate through OAuth in a browser. The package can also be run locally with an API key. A domain value can separate memories into namespaces such as user:<id>; this keeps data apart inside an account but is not a permission boundary.

memory_read returns matches ordered by relevance, or newest-first when requested with recency ordering. memory_list_recent provides a chronological catch-up path without a search query. It supports a since timestamp and cursor pagination; when reading a shared room, its address must be supplied as the domain. memory_stats reports counts, learned concept associations, available domains, and average quality scores.

Shared rooms provide separate collaborative stores. An owner creates a room, generates an invite, and can allow several people to use one invite. A recipient joins with an invite code and receives the room address. Membership scope determines whether that participant can write, while read-only members cannot rate room memories.

Setup and configuration

Create a Mnemoverse account and connect an MCP client to the hosted endpoint. Claude Code, Cursor, VS Code, Claude Desktop, Cline, Devin Desktop (formerly Windsurf), Zed, JetBrains, Gemini CLI, and ChatGPT are listed as supported setup targets. Clients with native remote MCP and OAuth support can use the URL directly. Other clients can use the mcp-remote bridge through npx; the bridge requires Node.js 20.18.1 or later and stores its sign-in data under ~/.mcp-auth.

For a local deployment of the MCP package, configure an API key instead of using hosted browser authentication. The memory engine itself remains a hosted service by default; Enterprise self-hosting is available by agreement for security or compliance requirements.

The free plan requires no credit card and includes a daily query allowance, storage capacity, and a per-minute request limit. Plan limits are published on Mnemoverse's pricing page.

Tools and capabilities

  • memory_write: Save durable context, optionally within a domain.
  • memory_read: Search memories by natural-language query and receive ranked results.
  • memory_list_recent: Page through newest memories, including entries added since a given time.
  • memory_feedback: Adjust ranking based on whether recalled memories helped or were stale.
  • memory_stats: Inspect counts, domains, associations, and quality scores.
  • memory_create_room, memory_invite_to_room, memory_join_room, and memory_list_rooms: Create and manage shared memory spaces.
  • vault_list: See secret aliases and purposes without exposing secret values.
  • memory_graph: Inspect concept links, weights, outcome valence, and co-activation counts.

Limitations and notes

The server is hosted by default, unlike a local-first memory store. Namespace separation with domain does not create an account-level permission boundary. Shared-room access depends on membership scope, and a read-only member cannot write or provide feedback for that room.

vault_list never returns secret values, and this server should not be used to store sensitive categories such as credentials or health records. Graph expansion at depth two or three filters low-weight edges unless a minimum weight is supplied. The repository and MCP server are MIT licensed, while the memory engine is operated as a hosted service.

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 MCP Memory Server

mnemoverse/mcp-memory-server is an MCP server that connects AI clients to the hosted Mnemoverse memory engine. Its main tools write and read durable memories, provide helpfulness feedback, list recent entries, inspect statistics, manage shared rooms, list secret aliases without values, and examine learned concept associations across connected tools.

Connect an MCP client to https://mcp.mnemoverse.com/mcp and complete the browser-based OAuth sign-in. Clients without native remote MCP support can use the mcp-remote bridge through npx. The package can also run locally with an API key.

Yes, Cursor is supported through the hosted Mnemoverse endpoint. Add the remote server URL in Cursor's MCP configuration, then use Connect and approve the browser sign-in. Cursor stores the authentication token itself, so no API key is required for this setup.

Shared rooms let multiple agents use one separate memory store. Create a room, mint an invite, and have another person join with the invite code. Use the returned room address as the domain for reads and writes; read-only members cannot write or rate room memories.

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 PreviewMCP Memory Server AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/mnemoverse-mcp-memory-server?style=directory)](https://allmcps.com/mcp/mnemoverse-mcp-memory-server)
HTML Embed
<a href="https://allmcps.com/mcp/mnemoverse-mcp-memory-server"><img src="https://allmcps.com/api/badge/mnemoverse-mcp-memory-server?style=directory" alt="MCP Memory Server on AllMCPs" /></a>