AutoGineer
Generated banner for Mnemoverse Memory
Generated logo for Mnemoverse Memory

Mnemoverse Memory

Memory serverOfficial Mnemoverse project

Created by Mnemoverse

Published in mnemoverse/mcp-memory-server. Publisher page

28 GitHub stars

What it is

Hosted long-term memory for AI agents over MCP. Write and read memories from Claude Code, Cursor, VS Code and ChatGPT; rating a recalled memory as helpful or misleading re-ranks what comes back next. OAuth sign-in or API key, free plan.
— the description by Mnemoverse

AutoGineer overview, from the brain’s own files

Mnemoverse Memory is a hosted memory server that enables AI agents to persist information across different sessions and tools. By utilizing the Model Context Protocol (MCP), it allows agents in environments like Claude, Cursor, and VS Code to read and write to a centralized memory store. The system features a feedback mechanism where agents can rate recalled memories, which the server uses to re-rank future retrieval results. It also supports collaborative memory through shared rooms, allowing multiple users or agents to contribute to a common knowledge pool.

How it works

Memory model: Hybrid

AutoGineer overview, from the brain’s own files

The system acts as an MCP server that provides tools for agents to interact with a hosted memory database. Agents use memory_write to store information and memory_read to retrieve it using natural language. The server maintains a graph of associations between concepts. Users can create shared rooms for collaborative memory, and agents can provide feedback on the utility of retrieved information, which influences the ranking of future search results.

Folders
FolderRoleFiles
.githubagent-config8
docsdocs24
scriptsscripts11
srcsource-code15
testtests49

The agent reads

The agent writes

Requirements

Runtimes: Node.js >=18

Databases: PostgreSQL

API keys (by name): MNEMOVERSE_API_KEY

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Google Antigravity, Kiro, OpenCode, Amp, Roo Code, Cline, JetBrains Junie, Devin (ex-Windsurf)

Set it up

  1. Add mcp-memory-server to .mcp.json:

    {
  "mcpServers": {
    "mcp-memory-server": {
      "type": "http",
      "url": "https://mcp.mnemoverse.com/mcp"
    }
  }
}

  2. Or add it with the Claude Code CLI:

    claude mcp add --transport http mcp-memory-server https://mcp.mnemoverse.com/mcp

  3. Your data is sent to mcp.mnemoverse.com.

  4. Add mnemoverse to .mcp.json:

    {
  "mcpServers": {
    "mnemoverse": {
      "type": "http",
      "url": "https://mcp.mnemoverse.com/mcp"
    }
  }
}

  5. Or add it with the Claude Code CLI:

    claude mcp add --transport http mnemoverse https://mcp.mnemoverse.com/mcp

  6. Your data is sent to mcp.mnemoverse.com.

  7. Not scanned by AutoGineerIts dependencies are installed by npm when you run it and were not scanned by AutoGineer (2 declared). https://www.npmjs.com/package/@mnemoverse/mcp-memory-server

  8. Created by Mnemoverse. Published in mnemoverse/mcp-memory-server. Licence: MIT License.

Usage

AutoGineer overview, from the brain’s own files

  • Store project-specific preferences or coding patterns for future sessions.
  • Retrieve historical context when starting work on a previously touched project.
  • Collaborate with teammates by sharing a memory room.
  • Provide feedback on memory relevance to improve future retrieval accuracy.
“Save that we prefer using Vitest for testing in this project.”
The agent writes this preference to the memory server, making it available for future tasks.
“How do we usually handle authentication in this codebase?”
The agent queries the memory server and retrieves previously saved notes on authentication patterns.

Documentation by Mnemoverse

Read the documentation at the source

AGENTS.md

AGENTS.md

Instructions for AI coding agents working in this repository. The human-facing version of the same rules is CONTRIBUTING.md; where the two differ, CONTRIBUTING.md wins.

If you are here to connect memory for a user

You do not need to change this repository. Use one of these:

  • ChatGPT: install the published Mnemoverse Memory plugin and complete browser sign-in. Start with https://mnemoverse.com/docs/api/chatgpt; no local package or custom MCP setup is needed for this route.

  • Hosted endpoint, no key to paste: https://mcp.mnemoverse.com/mcp (Streamable HTTP, OAuth sign-in in the browser). Per-client setup: https://mnemoverse.com/docs/api/install-by-host

  • Local stdio server: follow llms-install.md, which is written for an agent installing on a user's behalf. Have the user enter their key in the client's protected settings or a private user-level configuration, not in chat. Never invent a key or write a literal mk_live_ key into a file that gets committed.

Setup and checks

  • Node.js 18 or newer. Install with npm ci.

  • Build: npm run build (the prebuild step runs the config generator, the postbuild step regenerates tools.json).

  • Tests: npm test (vitest). Test types: npm run typecheck:test.

  • Drift checks: npm run verify:configs (generated files match src/configs/source.json) and npm run verify:tools (committed tools.json matches the built server).

CI runs the tests in three time zones, so do not assume UTC in tests that touch timestamps.

The one rule: src/configs/source.json

src/configs/source.json is the only file edited by hand for distribution metadata: descriptions, install snippets, env variables, registry fields. After changing it:

  1. npm run generate:configs

  2. npm run verify:configs

  3. Commit the source change together with every regenerated file.

Do not edit generated files by hand. They are server.json, manifest.json, everything under docs/configs/ and docs/snippets/, and the README regions between the INSTALL_SNIPPETS, MORE_CLIENTS and LOCAL_KEY_SNIPPETS HTML comment markers. A hand edit is overwritten on the next run and fails verify-configs.yml in the meantime. If a diff appears in one of these files and source.json did not change, discard it unless it is a regenerated tool-description change in manifest.json: the generator reads those descriptions from tools.json, which is built from src/tools.ts. For tool-description changes, run npm run build and commit the source plus regenerated tools.json and manifest.json.

Do not hard-code distribution metadata in src/, and do not add a parallel config file next to source.json.

Tool surface and changelog

Any change to what a tool does, returns or says about itself, including a description or parameter text, needs an entry under ## [Unreleased] in CHANGELOG.md in the same PR. That text is what a model reads, so even a wording change is a release-worthy patch. An internal refactor with identical behaviour gets [skip changelog] in the PR description instead. changelog-reminder.yml enforces this for changes under src/.

Versions and releases

The version lives only in package.json; src/index.ts reads it and server.json is generated from it. Releases are cut by pushing a vX.Y.Z tag, and .github/workflows/release.yml publishes to npm and the MCP Registry. Do not bump versions, tag or publish unless a maintainer asks you to.

After an authorized package release, follow the wave tracker and deployment checks. A successful publication does not prove the docs or hosted connector updated, and does not authorize an otherwise excluded production deployment.

Pull requests

Keep one concern per PR, run npm test and npm run verify:configs before pushing, and say in the description which of the checks above you ran.

134 files · 124 not served here

Advantages

Limitations

Security

Risk level: Medium risk

This brain runs

Agent instructions that run or fetch something (14)

  • AGENTS.md: ChatGPT: install the published Mnemoverse Memory plugin and complete browser sign-in.
  • AGENTS.md: Per-client setup: https://mnemoverse.com/docs/api/install-by-host
  • AGENTS.md: Local stdio server: follow `llms-install.md`, which is written for an agent installing on a user's behalf.
  • AGENTS.md: Install with `npm ci`.
  • AGENTS.md: Build: `npm run build` (the prebuild step runs the config generator, the postbuild step regenerates `tools.json`).
  • AGENTS.md: Tests: `npm test` (vitest).
  • AGENTS.md: Test types: `npm run typecheck:test`.
  • AGENTS.md: Drift checks: `npm run verify:configs` (generated files match `src/configs/source.json`) and `npm run verify:tools` (committed `tools.json` matches the built server).
  • AGENTS.md: `src/configs/source.json` is the only file edited by hand for distribution metadata: descriptions, install snippets, env variables, registry fields.
  • AGENTS.md: `npm run generate:configs`
  • AGENTS.md: `npm run verify:configs`
  • AGENTS.md: A hand edit is overwritten on the next run and fails `verify-configs.yml` in the meantime.
  • AGENTS.md: For tool-description changes, run `npm run build` and commit the source plus regenerated `tools.json` and `manifest.json`.
  • AGENTS.md: Keep one concern per PR, run `npm test` and `npm run verify:configs` before pushing, and say in the description which of the checks above you ran.

MCP servers (11)

  • README.md (mnemoverse): npx -y mcp-remote@latest https://mcp.mnemoverse.com/mcp
  • README.md (mnemoverse): npx -y mcp-remote@latest https://mcp.mnemoverse.com/mcp
  • README.md (mnemoverse): https://mcp.mnemoverse.com/mcp · uses the network
  • README.md (mnemoverse): npx -y mcp-remote@latest https://mcp.mnemoverse.com/mcp
  • README.md (mnemoverse): npx -y mcp-remote@latest https://mcp.mnemoverse.com/mcp
  • README.md (mnemoverse): npx -y @mnemoverse/mcp-memory-server@latest
  • README.md (mnemoverse): npx -y @mnemoverse/mcp-memory-server@latest
  • README.md (mnemoverse): npx -y @mnemoverse/mcp-memory-server@latest
  • README.md (mnemoverse): npx -y @mnemoverse/mcp-memory-server@latest
  • README.md (mnemoverse): npx -y @mnemoverse/mcp-memory-server@latest
  • gemini-extension.json (mnemoverse): https://mcp.mnemoverse.com/mcp · uses the network
  • Medium risk: runs-mcp-server-unpinned-runner (README.md:60)
  • Medium risk: runs-mcp-server-unpinned-runner (README.md:100)
  • Medium risk: runs-mcp-server-unpinned-runner (README.md:132)
  • Medium risk: runs-mcp-server-unpinned-runner (README.md:171)
  • Medium risk: runs-mcp-server-unpinned-runner (README.md:275)
  • Medium risk: runs-mcp-server-unpinned-runner (README.md:324)
  • Medium risk: runs-mcp-server-unpinned-runner (README.md:344)
  • Medium risk: runs-mcp-server-unpinned-runner (README.md:364)
  • Medium risk: runs-mcp-server-unpinned-runner (README.md:405)
  • Low risk: network-instructions (package.json:49)
  • Low risk: scripts-present (scripts/generate-configs.mjs)

Attribution and licence

Created by
Mnemoverse
Declared authors
  • Mnemoverse (declared in package.json)
Publisher
Mnemoverse (Official Mnemoverse project)
Source
mnemoverse/mcp-memory-server · commit 9fd95f6
Licence
MIT License (files served here)
  • ATTRIBUTION.md (written by AutoGineer)
Synced

AutoGineer is not affiliated with or endorsed by Mnemoverse.

Install from source

124 of its files are not served here, so there is no ZIP. mnemoverse/mcp-memory-server at 9fd95f6