Skip to main content
Glama
ledgermem

LedgerMem MCP Server

Official
by ledgermem
OverviewSchemaSourceRelated ServersDiscussionsAdmin
TypeScript
Hybrid

getmnemo-mcp

Model Context Protocol server for Mnemo Memory — exposes long-term memory tools to any MCP client (Claude Desktop, Cursor, Windsurf, VS Code, Zed).

Tools

Tool

What it does

memory_search

Hybrid 7-strategy retrieval over the workspace memory store. Filter by polarity, searchMode (memories/documents/both), exclude already-seen ids, or opt into precise mode.

memory_answer

Cited-answer pipeline: question in, synthesized answer + citations out (mode: "fast" for ~1-2s).

memory_add

Store an atomic fact with optional metadata; polarity: "negative" marks a hard constraint.

memory_get

Fetch one memory by ID within the configured container.

memory_update

Patch an existing memory's content, metadata, or polarity.

memory_delete

Soft-delete a memory by ID so it leaves retrieval.

memory_restore

Undo a soft delete during its recovery window (API-key sessions only).

memory_list

Paginate through memories (cursor-based).

document_add

Ingest a raw document (up to 500,000 characters) into the async extraction lane; returns a jobId.

job_status

Poll an ingestion job: queued / processing / completed / failed (API-key sessions only).

Personal-memory tools (0.3.0, Mnemo API v0.3.0)

Tool

What it does

Key scope(s)

API flag

daily_brief

Today's reminders, important dates, last-24h captures, follow-ups and meetings for one container.

brief:read

MEMORY_API_BRIEF_ENABLED

memory_timeline

One container as a time-ordered stream of memories, reminders, documents (and optional events).

timeline:read

MEMORY_API_TIMELINE_ENABLED

people_list

People the workspace remembers, with memory / open-reminder counts.

people:read

MEMORY_API_PEOPLE_ENABLED

people_get

One person by slug (contact details, important dates, aliases).

people:read

MEMORY_API_PEOPLE_ENABLED

people_upsert

Create a person, or update them when the slug already exists.

people:write

MEMORY_API_PEOPLE_ENABLED

reminder_create

A reminder with a due time, filed under a person or a container.

reminders:write

MEMORY_API_PEOPLE_ENABLED

reminders_upcoming

Overdue / due-today / upcoming reminders plus important dates, by local day.

reminders:read

MEMORY_API_PEOPLE_ENABLED

reminder_complete

Mark a reminder done.

reminders:write

MEMORY_API_PEOPLE_ENABLED

meetings_upcoming

Upcoming Google Calendar meetings with attendees resolved to people.

meetings:read

MEMORY_API_MEETINGS_ENABLED

meeting_brief

Pre-meeting brief: reader answer with citations, matched people, previous meetings.

meetings:read + answer:read

MEMORY_API_MEETINGS_ENABLED

memory_merge

Fold 2..20 duplicate memories into one survivor (idempotent on mergeKey).

memories:write + memories:delete

MEMORY_API_INBOX_ENABLED

These scopes are opt-in: mint (or edit) the API key with them at https://app.mnemohq.com/settings/api-keys. Every feature is dark until the API operator sets its MEMORY_API_* flag; until then the tool returns a 503 FEATURE_DISABLED error that names the flag.

All personal-memory tools except memory_timeline are API-key only: the API rejects hosted OAuth MCP tokens on cross-container routes, so hosted OAuth sessions (https://mcp.mnemohq.com/mcp) do not list them.

Remote HTTP calls are scoped by the user's Mnemo OAuth grant. The user signs in, chooses a workspace and memory container, and the server derives that scope from the access token. The model cannot set or change it.

Related MCP server: Memorium

Install

Local stdio clients

npx -y getmnemo-mcp

Or wire it into the client config directly. Example for Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "getmnemo": {
      "command": "npx",
      "args": ["-y", "getmnemo-mcp"],
      "env": {
        "GETMNEMO_API_KEY": "prfly_live_...",
        "GETMNEMO_WORKSPACE_ID": "ws_...",
        "GETMNEMO_CONTAINER_TAG": "user:jane"
      }
    }
  }
}

Get an API key at https://app.mnemohq.com/settings/api-keys.

Hosted remote MCP clients

For clients that support remote MCP and OAuth 2.1, use:

https://mcp.mnemohq.com/mcp

The client should discover OAuth metadata, open Mnemo sign-in, and return to the client after the user selects a workspace and container. No API key, workspace header, or container configuration is required for this path.

Develop

npm install
cp .env.example .env   # fill in API-key/container values for stdio
npm run dev            # stdio
npm run dev:http       # Streamable HTTP on :8787
npm run build          # bundle to dist/
npm run lint && npm run typecheck && npm test

Architecture

  • stdio (src/cli.ts): one process per MCP client connection, env-configured.

  • Streamable HTTP (src/http.ts): hosted transport for public OAuth sessions, with private API-key compatibility for trusted deployments.

  • Both transports share src/server.ts (tool registration + dispatch), src/personal-tools.ts (personal-memory tools) and src/api-client.ts (typed REST wrapper).

The public hosted transport uses Mnemo's OAuth 2.1 authorization server and resource-bound tokens. Header-selected containers remain disabled by default; they are only a private compatibility path for a trusted gateway.

The server deliberately does NOT depend on getmnemo (the JS SDK) so it can ship independently.

License

MIT

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityMaintained
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers