Skip to main content
Glama

MAGNEMO is Memory for your AI, kept on your computer in plain files. Your AI writes down what it learns as it works; nothing becomes Memory until you approve it, and every Memory shows who wrote it and who approved it.

Your agent writes one line down and keeps working. Say keep, and it is Memory.

PyPI downloads tests license stars MCP registry

Site · Manual · Changelog · X @Magnemo_AI · Known limits

Install in one paste

In three lines:

pip install MAGNEMO      # Python 3.11 or newer
magnemo init             # makes the vault: a folder of plain files
magnemo mount            # writes .mcp.json, the file Claude Code reads, and the first line in CLAUDE.md

Or let your agent do it. Open Claude Code in any of your project folders, and paste this. (Its magnemo mount step writes .mcp.json, the file Claude Code reads, and a short block in CLAUDE.md that tells every new session where this project's Memory is. In Cursor, Windsurf or another client, use the block for your client under "Install in your client".)

Set up MAGNEMO in this project:
1. Install it: pip install MAGNEMO. It needs Python 3.11 or newer. If pip can't find it, install uv (curl -LsSf https://astral.sh/uv/install.sh | sh) and then run: uv tool install MAGNEMO
2. Initialize the vault: magnemo init
3. Register the MCP server: run magnemo mount — it adds MAGNEMO to this project's MCP config and keeps any servers already there.
4. Verify: run magnemo doctor and confirm everything passes.
5. Then ask me to restart you, so MAGNEMO's tools load, and to say this when you are back: "Look around this folder, tell me what this project is for, and stage that as the first note." Your agent never waits on you. When you want something kept, say keep.
If anything fails, show me what went wrong and fix it if you can.

Plant the seed: one paste, and your agent grows a Memory.

After the restart, say it: Look around this folder, tell me what this project is for, and stage that as the first note. Your agent writes one line down and keeps working. Say keep, and it is Memory. Then ask what it remembers.

Related MCP server: Engram

Works with

One server, the same in every client. Tested where it says so; the rest follow each client's own docs.

Client

Tested

How

Claude Code

yes: Oct 26–28, 2026, Claude Code 2.1.309. Seven tools; the yes in the chat; pending reads a body.

magnemo mount, or claude mcp add magnemo -e MAGNEMO_VAULT=/absolute/path/to/vault -- uvx magnemo-mcp

Claude Desktop

untested end to end

the uvx block under "Install in your client", with absolute paths

Cursor

untested

the Add to Cursor link below, or .cursor/mcp.json

VS Code

untested

.vscode/mcp.json

Windsurf

untested

~/.codeium/windsurf/mcp_config.json

Cline

untested

cline_mcp_settings.json

Continue

untested

~/.continue/config.yaml

Gemini CLI

untested

~/.gemini/settings.json

Codex CLI

untested

~/.codex/config.toml

Zed

untested

settings.json → context_servers

Each client's exact block is under "Install in your client".

How it compares

Five questions, answered from each project's own docs (read Oct 12, 2026, checked again Oct 22; the pages are linked under the table).

MAGNEMO

Mem0

Zep

Letta

Basic Memory

Where the memory lives

Plain markdown files in a folder on your computer

A hosted platform run by Mem0; open source: a local Qdrant store and a SQLite history file

Zep: a managed cloud service, or in your own VPC. Graphiti, its open-source engine: a graph database you run

Markdown files in a git repository that belongs to the agent: hosted by Letta, or on your device in local mode

Markdown files on your disk with a local SQLite index; an optional cloud

Who approves a write

You. An agent's write waits in staging until your yes

An LLM extracts facts from the messages and stores them; their docs describe no approval step

Messages are ingested into the graph as they arrive; their docs describe no approval step

The agent updates its own memory; the background review "does not ask you for approval"

The assistant writes notes directly; the cloud has suggested edits that can be accepted or rejected

Author and approver on every entry

Both, in each file's header (author, reviewed_by) and in an append-only ledger

User, agent, app and run ids, timestamps, and a change history per memory; their docs name no approver field

When a fact was true and when it stopped, traced to its source episode; their docs name no approver field

Every memory edit is a git commit; their docs do not say what the commit records

Title, type, tags and dates; author is a field you may add; their docs name no approver field

Network calls of its own

None. No model call and no embedding call; git push to a backup, if you add one

OpenAI by default for the model and the embeddings; other providers are listed, Ollama among them

Zep is a cloud API. Graphiti defaults to OpenAI; local model servers are documented

Local mode keeps state on your device; inference goes to the provider you choose, which can be a local one

Runs locally; the default embeddings run on your machine with no API key

What you can take back

A no is kept with its reason; a replaced note is archived with a forward link; nothing deletes; the files are yours to edit

Delete, update, and a history per memory; export is on the hosted platform

Delete nodes, edges and episodes; a replaced fact is invalidated, not deleted

The git history of every memory edit; the files can be read and edited

Plain files you edit and can track with git; a delete tool; the cloud keeps every saved version

Where each of them is ahead. Mem0 extracts memories for you, searches them by meaning, and has Python and JavaScript SDKs, a REST API and a hosted platform. Zep keeps a knowledge graph that knows when each fact was true, as a managed service with SDKs in three languages. Letta is a whole agent platform, with an SDK, a CLI and background agents that tend the memory. Basic Memory searches by keyword and by meaning with embeddings that run on your machine, links notes into a graph, and syncs through its cloud. MAGNEMO has none of that: its search is keyword search (KNOWN_LIMITS.md), there is no graph and no hosted service. What it has is the approval and the record.

Their pages: Mem0 overview, platform and open source, add, history, quickstart, delete, export · Zep concepts, facts, security, deleting data, Graphiti · Letta MemFS, memory settings, local mode · Basic Memory README, what it is, tools, note format, semantic search, suggestions, file history. A cell that is wrong or out of date: open an issue and it is fixed.

Listed on the official MCP registry (ai.magnemo/magnemo) · PyPI · mcpservers.org · Glama · awesome-mcp-servers. Tested on every push, on Python 3.9, 3.11 and 3.12: 370 tests, as of Oct 28, 2026 (the runs).

What's new

  • 0.6.11 · Oct 28, 2026 · magnemo mount writes the first line in CLAUDE.md, so every new session knows where this project's Memory is. pending reads a note nobody has kept yet.

  • 0.6.10 · Oct 26, 2026 · Your agent never stops to ask. Say keep in the chat, and the record says how the yes was given.

  • 0.6.9 · Oct 22, 2026 · A seat's report files itself as a receipt and asks nothing; a yes is for Memory. Sentinel reads a seat's quote as a quote.

Every release: CHANGELOG.md.

For builders. MAGNEMO is an MCP server with four verbs — retrieve, stage, bootpack, handoff — plus keep, drop, pending, and save and rewind, and a CLI for your side: review, promote, reject. It runs on your machine, in plain markdown files you own. The package imports no network module (magnemo doctor scans for one on every run); the only traffic is git push to a backup destination, if you add one. Every promotion is recorded with who wrote the entry, the name given for the yes, when, and through which gate. Trust is computed from that record, never asserted. It mounts beside any memory you already run (a folder of notes, Obsidian, a RAG stack). Python 3.11+, zero dependencies, Apache-2.0.

Fresh vault? THE CHARTER section will say it isn't promoted yet — that is the governance speaking: MAGNEMO never fabricates canon. Stage a charter, review it, promote it, and every boot thereafter serves it back verbatim. (magnemo not found after install? pip's script folder isn't on PATH — python3 -m magnemo always works.)

Open beta — read KNOWN_LIMITS.md before trusting it with anything you can't lose. Honest caveats, no fine print.

Install in your client

One server, said ten ways. Every block below is the same thing: uvx magnemo-mcp with MAGNEMO_VAULT pointing at the folder your Memory lives in (make one with magnemo init ./vault). Tested on this Mac where it says so; the rest is from each client's own docs and marked untested — corrections welcome, open an issue.

Claude Code — tested: a Claude Code session mounts the server, lists the seven tools and calls them; a yes in the chat keeps a note, and pending reads a body (last run Oct 28, 2026: Claude Code 2.1.309, the 0.6.11 server).

claude mcp add magnemo -e MAGNEMO_VAULT=/absolute/path/to/vault -- uvx magnemo-mcp

(claude mcp list shows magnemo … ✔ Connected. Add -s project to write it into the repo's .mcp.json for your team.) Only have the desktop app? claude may not be a command in your shell. On a Mac the app keeps its own copy at ~/Library/Application Support/Claude/claude-code/<version>/claude.app/Contents/MacOS/claude; run that in place of claude.

Claude Desktop — untested end to end here. The server this block starts was run the way Claude Desktop starts it (Nov 8, 2026, the 0.6.13 server). Claude Desktop → Settings → Developer → Edit Config, then claude_desktop_config.json. Claude Desktop does not read your shell's PATH, so both paths are absolute: which uvx prints the first.

{
  "mcpServers": {
    "magnemo": {
      "command": "/absolute/path/to/uvx",
      "args": ["magnemo-mcp"],
      "env": { "MAGNEMO_VAULT": "/absolute/path/to/vault", "MAGNEMO_AGENT": "claude-desktop" }
    }
  }
}

Restart Claude Desktop after saving.

Cursor — untested here (not installed on this Mac); the link and the block follow Cursor's docs. Add to Cursor Or ~/.cursor/mcp.json (or .cursor/mcp.json in the project):

{
  "mcpServers": {
    "magnemo": {
      "command": "uvx",
      "args": ["magnemo-mcp"],
      "env": { "MAGNEMO_VAULT": "/absolute/path/to/vault" }
    }
  }
}

VS Code / Copilot — untested here; per VS Code's docs. Install in VS Code Or .vscode/mcp.json in the workspace:

{
  "servers": {
    "magnemo": {
      "type": "stdio",
      "command": "uvx",
      "args": ["magnemo-mcp"],
      "env": { "MAGNEMO_VAULT": "/absolute/path/to/vault" }
    }
  }
}

Windsurf — untested; ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "magnemo": {
      "command": "uvx",
      "args": ["magnemo-mcp"],
      "env": { "MAGNEMO_VAULT": "/absolute/path/to/vault" }
    }
  }
}

Cline — untested; Cline → MCP Servers → Configure → cline_mcp_settings.json:

{
  "mcpServers": {
    "magnemo": {
      "command": "uvx",
      "args": ["magnemo-mcp"],
      "env": { "MAGNEMO_VAULT": "/absolute/path/to/vault" },
      "disabled": false
    }
  }
}

Continue — untested; ~/.continue/config.yaml (or a file in .continue/mcpServers/):

mcpServers:
  - name: magnemo
    command: uvx
    args: ["magnemo-mcp"]
    env:
      MAGNEMO_VAULT: /absolute/path/to/vault

Gemini CLI — untested; ~/.gemini/settings.json:

{
  "mcpServers": {
    "magnemo": {
      "command": "uvx",
      "args": ["magnemo-mcp"],
      "env": { "MAGNEMO_VAULT": "/absolute/path/to/vault" }
    }
  }
}

Codex CLI — untested; ~/.codex/config.toml:

[mcp_servers.magnemo]
command = "uvx"
args = ["magnemo-mcp"]
env = { MAGNEMO_VAULT = "/absolute/path/to/vault" }

Zed — untested; settings.json → context_servers:

{
  "context_servers": {
    "magnemo": {
      "source": "custom",
      "command": "uvx",
      "args": ["magnemo-mcp"],
      "env": { "MAGNEMO_VAULT": "/absolute/path/to/vault" }
    }
  }
}

After the install, the loop is the same everywhere: your agent writes to staging, you review.

magnemo yes             # promote the top of the queue (yes <id-fragment> · yes --all)
magnemo no <id> --reason "…"   # reject one — the reason is required; rejections teach
magnemo review          # the interactive queue, when you want to read first
magnemo doctor          # python, vault, config, ledger, and the mount — scans the package for network modules too

Anything that speaks MCP over stdio mounts the same way. The remote door (ChatGPT and hosted clients) is on its way in 0.7.0.

How it works

The yes is checked: nothing becomes Memory without a yes, and the engine can tell a person's yes from an agent's; a yes in the chat counts, and the record says how it was given.

MAGNEMO is Memory with a constitution: every fact carries provenance, every agent write waits in your review queue before it becomes Memory, and every decision lands in a Trust Ledger.

Your agent gets four verbs and cannot keep a note on its own authority; it carries your yes, and the record says so. Say yes in the chat and keep carries it, ledgered unverified; a click your client draws or a yes at your terminal is a verified one. KNOWN_LIMITS.md says what that does and does not stop; docs/MCP.md has the clients it works in.

Actor

Door

Can do

Agents

MCP server (stdio)

retrieve canonical memory · stage → staging only, the one write tool · bootpack on wake · handoff at the boundary · keep / drop carry your yes or no · pending lists what waits

You

CLI + git (any editor)

review the queue · promote / reject · edit anything · own everything

Zero dependencies. Pure Python standard library, 3.11 or newer. Nothing else to install, nothing to trust but this code.

The vault

vault/
  dev/      knowledge/ playbooks/ decisions/ debt/        ← example: a dev team's partition
  ops/      knowledge/ playbooks/ clients/ decisions/ style/  ← example: an ops team's partition
  shared/   tickets/ changelog/                            ← interop bus (gated)
  _staging/   agent writes await review here
  _index/     machine-managed
  _ledger/    trust_ledger.jsonl — append-only, never forgets

Every note is markdown with provenance frontmatter (author, written, source, status, reviewed_by, supersedes, strength). Provenance is the file format.

Privacy

Your Memory lives on your computer, in plain files you can open. No network calls of its own. Add a backup remote and it pushes there, nowhere else. The precise version: the package imports no network module, and magnemo doctor scans for one every time it runs. If you add a backup destination, the engine runs git push to it. The full page: https://magnemo.ai/privacy

Guarantees (Phase 1 — Governed Recall)

  • Through the MCP server, agent writes NEVER reach canonical stores directly — staging only, always.

  • Search returns canonical (promoted) notes only. Staged notes are not searched. The boot pack shows staged material marked as staged: the titles of open threads and of the top staged notes, and the last handoff's cut line with a pointer to its note.

  • Partition walls enforced per agent (MAGNEMO_PARTITIONS); cross-partition = DENIED.

  • Supersession is explicit: old notes archive with a forward link. Nothing deletes.

  • Rejections are kept and recorded — rejections teach.

  • Every promote/reject lands in the append-only Trust Ledger with actor + reason.

Tests

python -m unittest discover -s tests -v     # incl. full MCP round-trips (legacy + four-verb)
# or: pip install -e ".[test]" && pytest

The same suite runs on every push, on Python 3.9, 3.11 and 3.12: the runs.

Everything that shipped

Release by release, with the design notes that used to live here: CHANGELOG.md.

— Silver Valley Technologies Inc. · Phase 1 of 3 · The memory that learns is the memory that is governed.


mcp-name: ai.magnemo/magnemo · Source · Security · Privacy

Available Tools

7 tools
bootpackRead what this agent wakes up knowingA
Read-onlyIdempotent

The boot pack: Charter first (verbatim, never fabricated), then the Gate Map, YOUR AUTONOMY (this agent's own trust levels), canon digest, open threads, salience-ranked review queue, last handoff, ledger tail. Read it FIRST on every wake. Deterministic: same vault state, same bytes.

Inputs: class, scope.

dropDrop a note the person said no toA
Idempotent

Call drop only when the person has said no to this note in their own words in this conversation. The note is set aside as rejected, with their reason, and never becomes Memory.

Inputs: note (required), reason.

handoffRecord where this session stoppedA

Record a session boundary: appends telemetry (usage %, trigger, what was cut) to the append-only handoffs ledger AND stages a provenance-complete handoff note. The next bootpack inherits it. Doctrine: hand off at ~88-90%, do not ride to the wall. A pause: where the client can draw a question, the person is asked once about every note still staged (keep all / pick / later); later or no answer leaves them staged and blocks nothing.

Inputs: usage (required), trigger (required), cut.

keepKeep a note the person said yes toA
Idempotent

Call keep only when the person has said yes to this note in their own words in this conversation; never draw a panel, and never keep on your own judgment. The note becomes Memory, and the record says the yes was given in the chat (unverified). A note Sentinel is holding is not kept this way. "all" keeps every note pending lists, when the person said keep all; each is its own ledger line. The cap: at most 50 notes a day are kept by a yes in the chat, counted in the local day of the computer the vault is on and reset at its midnight. Two ways past it: the person's own terminal (magnemo yes ), or the one question their client draws at a pause (handoff).

Inputs: note (required).

pendingList what waits for the person's keepA
Read-onlyIdempotent

The staged notes nobody has answered yet, and the one line to say about them. Call it at a pause, at the end of your own turn, never mid-work; say its line once ('N notes staged since you left. keep all / pick / later.'). Later, or no answer, is fine: nothing blocks and nothing is lost. With a note's id, it returns that staged note's full body and where it came from, labelled 'staged · not yet Memory' (a note Sentinel holds gives its alert, not its body).

Inputs: note.

retrieveLook something up in memoryA
Read-onlyIdempotent

Retrieve canonical (keyholder-approved) memory relevant to a query. Returns a scored, BUDGETED payload: mn:// pointers plus snippet renditions, never raw dumps. Scope is partition-walled; payload size is logged to the Foresight counters. Cite the returned rid when the action resolves.

Inputs: query (required), k, scope, budget.

stageDraft an entry for a person to reviewA

Remembering means proposing: this stages a note; the person's keep makes it Memory. Contact details are kept as written and labelled; the label is checked once before anything leaves this computer. Stage a memory candidate for the person's review. This is the only write verb: the note lands in _staging/ with mandatory provenance and never touches canon. It never waits and asks nothing: keep working. Returns the staged id, its salience, a duplicate flag, and one line to say: 'Staged. Say keep when you want it kept.' Ask about what you staged only at a pause (see pending), never mid-work. Malformed or paperless (no provenance) notes are refused.

Inputs: note (required), provenance (required).

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4/5.0

    Scored across 7 tools

    Disambiguation4/5

    Each tool has a clearly stated role in the memory lifecycle (stage/keep/drop/pending/retrieve) plus session tools (handoff/bootpack). Minor overlap exists: handoff also stages a note (overlapping with stage), and pending and bootpack both surface staged/review content, but the descriptions delineate their boundaries well enough.

    Naming Consistency3/5

    Names are all lowercase single tokens with no case-style inconsistencies, but they mix verbs (stage, keep, drop, retrieve) and nouns (handoff, pending, bootpack) with no predictable verb_noun pattern. Readable but not a consistent convention.

    Tool Count5/5

    Seven tools map cleanly onto the vault's capture-review-approve-retrieve-session lifecycle; each tool earns its place and none appears redundant filler.

    Completeness4/5

    Core lifecycle is covered: propose, approve, reject, list pending, retrieve canon, boot context, and session handoff. Gaps exist around editing/amending a staged note and managing canon directly (no update/delete of Memory), but agents can work around these via stage/drop.

    Maintenance

    ActivityActive
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      C
      maintenance
      A local-first MCP server that gives AI assistants long-term memory by storing, searching, and recalling notes as Markdown files on your machine.
      4 npm
      MIT
    • A
      license
      A
      quality
      C
      maintenance
      A self-hosted MCP server that gives AI agents shared, long-term memory over a git-backed folder of markdown, enabling persistent knowledge search, read, and write without a database.
      16
      23
      MIT