Skip to main content
Glama
La Marzocco MCP

La Marzocco MCP

by EmilbMadsen

lamarzocco-mcp

Unofficial. This project is not affiliated with, endorsed by, or supported by La Marzocco.

An MCP server that lets Claude and other MCP clients read the status, settings and shot data of La Marzocco home espresso machines, and optionally change a few settings. It runs locally over stdio and is built on pylamarzocco.

Ask things like "what was my last shot?", "is the machine on?" or, with write tools enabled, "turn on the machine".

Supported machines

Developed and tested on a Linea Micra. Other models supported by pylamarzocco should work; see docs/capabilities.md for what has been verified per model.

Related MCP server: AppleScript Automation MCP

Requirements

  • uv (it installs Python 3.13 for you if needed)

  • A La Marzocco account that uses a regular email/password login (not Google or Apple sign-in)

  • The machine connected to wifi and set up in La Marzocco's app

Setup

Store your La Marzocco login in the system keyring (macOS Keychain, Windows Credential Locker, or the Secret Service on Linux). Nothing is written to disk:

uvx lamarzocco-mcp setup

It signs in, lists your machines and saves the one to use. Run it again to change account, or uvx lamarzocco-mcp setup --delete to remove everything. On macOS, the first time the server reads the keychain you may be asked to allow it; choose "Always Allow".

Use with Claude

The server runs over stdio: your MCP client starts it when needed, so there is nothing to keep running.

Claude Code:

claude mcp add --scope user lamarzocco -- uvx lamarzocco-mcp

Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json on macOS), then restart Claude Desktop:

{
  "mcpServers": {
    "lamarzocco": {
      "command": "uvx",
      "args": ["lamarzocco-mcp"]
    }
  }
}

If Claude Desktop cannot find uvx, use its full path (which uvx, e.g. /opt/homebrew/bin/uvx).

Tools

Read-only by default. Each tool wraps one method of pylamarzocco's LaMarzoccoCloudClient.

Tool

Returns

list_machines

Machines on the account

get_machine_status

On/standby, boilers, pre-brewing, back flush

get_machine_settings

Firmware, wifi signal, plumb-in, auto-update

get_statistics

Statistics widgets from the app

get_firmware_status

Firmware update status

get_schedule

Smart standby and wake-up/sleep schedules

get_coffee_and_flush_counter

Total coffees and flushes

get_last_coffees

Latest 15 shots: time, extraction (pump) seconds, dose, temperature

get_coffee_and_flush_trend

Coffees and flushes per day, up to 30 days

Responses are La Marzocco's raw JSON with three changes: timestamps become ISO 8601 in local time, the Bluetooth token and wifi name are redacted, and image/tutorial URLs are dropped.

Shot times (extractionSeconds) are pump time measured by the machine, usually a few seconds shorter than a scale's timer.

Optional write tools

Off by default. With LAMARZOCCO_MCP_ALLOW_WRITE=1 in the server's environment, seven more tools are registered:

Tool

Changes

set_power

On (brewing mode) or standby

set_steam

Steam boiler on/off

set_steam_target_level

Steam level 1–3 (level-based machines such as the Linea Micra)

set_coffee_target_temperature

Coffee boiler target, °C

set_pre_extraction_mode

Pre-brewing on/off (or pre-infusion, where the machine has it)

set_pre_extraction_times

Pre-brewing seconds in and out

set_smart_standby

Automatic standby after N minutes

Each request is checked against the limits the machine reports before it is sent, and the setting is read back afterwards. Back flush, plumb-in, firmware updates and schedules are not exposed.

To enable them in Claude Desktop, add an env block to the server entry:

"lamarzocco": {
  "command": "uvx",
  "args": ["lamarzocco-mcp"],
  "env": { "LAMARZOCCO_MCP_ALLOW_WRITE": "1" }
}

In Claude Code: claude mcp add --scope user -e LAMARZOCCO_MCP_ALLOW_WRITE=1 lamarzocco -- uvx lamarzocco-mcp.

Configuration

Variable

Effect

LAMARZOCCO_MCP_ALLOW_WRITE=1

Registers the write tools

LAMARZOCCO_SERIAL

Use this machine instead of the one chosen in setup

LAMARZOCCO_CREDENTIALS_FILE

JSON file with username, password, installation_key (and optionally serial_number) instead of the keyring, e.g. a Docker secret

With several machines on the account, tools also take an optional serial_number.

Remote access (optional)

pip install 'lamarzocco-mcp[remote]' adds lamarzocco-mcp serve --http: Streamable HTTP at /mcp behind single-user OAuth 2.1 (password + TOTP, PKCE, short-lived rotating tokens) from mcp-owner-auth, so the server can be added as a custom connector in claude.ai and used from the Claude mobile app. Self-hosted and single-user only. A Dockerfile is included.

Variable / command

Effect

LAMARZOCCO_MCP_PUBLIC_URL

public base URL, e.g. https://lamarzocco.example.com

LAMARZOCCO_MCP_STATE_DIR

where auth.db and the audit log live (default ~/.config/lamarzocco-mcp)

LAMARZOCCO_MCP_CLIENT_IP_HEADER

e.g. CF-Connecting-IP behind a Cloudflare Tunnel

lamarzocco-mcp auth init

username, password and TOTP; prints the OAuth client ID and secret once

lamarzocco-mcp auth revoke-all

revokes every token immediately

Remote access is read-only: serve --http refuses to start with LAMARZOCCO_MCP_ALLOW_WRITE=1 unless LAMARZOCCO_MCP_ALLOW_REMOTE_WRITE=1 is also set.

Security

  • Your login and installation key are stored only in the system keyring (or, for a server, in the file LAMARZOCCO_CREDENTIALS_FILE points to).

  • By default the server runs only locally over stdio; there is no network listener. Remote access is opt-in (see above).

  • Read-only unless you enable write tools.

  • The server uses La Marzocco's undocumented cloud API through pylamarzocco, the same API their app uses. It can stop working if La Marzocco changes it.

Contributing

Have a different La Marzocco model? Mapping what it returns takes about ten minutes; see CONTRIBUTING.md.

Credits

License

MIT

Available Tools

9 tools
get_coffee_and_flush_counterB
Read-only

Total number of coffees and flushes the machine has made.

Inputs: serial_number. Returns: object

get_coffee_and_flush_trendA
Read-only

Coffees and flushes per day for the last N days, in the user's local timezone.

Inputs: days, serial_number. Returns: object

get_firmware_statusB
Read-only

Firmware update status (whether the machine is up to date or updating).

Inputs: serial_number. Returns: object

get_last_coffeesA
Read-only

The machine's most recent shots, newest first (La Marzocco keeps the latest 15).

Each shot has its end time, extractionSeconds, dose mode, target temperature and whether La Marzocco considers it valid. extractionSeconds is pump time measured by the machine, not a scale timer, so it is usually a few seconds shorter than a scale shows.

Inputs: limit, serial_number. Returns: object

get_machine_settingsC
Read-only

Machine settings: firmware versions, wifi signal, plumb-in, auto-update.

Inputs: serial_number. Returns: object

get_machine_statusA
Read-only

Current machine status: on or standby, coffee and steam boiler state and targets, pre-brewing, and back flush. From the dashboard in La Marzocco's app.

Inputs: serial_number. Returns: object

get_scheduleC
Read-only

Smart standby, wake-up/sleep schedules and early start settings.

Inputs: serial_number. Returns: object

get_statisticsC
Read-only

The statistics widgets selected in La Marzocco's app, e.g. the last coffees and the total coffee and flush counters.

Inputs: serial_number. Returns: object

list_machinesA
Read-only

List the La Marzocco machines on the account: model, name, serial number, online state.

Inputs: no named parameters. Returns: object

Tool Schema Changelog

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

Loading tool schema changelog…

    TDQS

    B3.4/5.0

    Scored across 9 tools

    Disambiguation3/5

    Most tools target distinct aspects of a machine, but get_statistics overlaps materially with get_coffee_and_flush_counter, get_last_coffees, and get_coffee_and_flush_trend by potentially returning the same counters or recent shots. get_firmware_status also overlaps somewhat with get_machine_settings, which includes firmware versions. Descriptions help, but an agent could still choose the wrong tool for a simple counter or recent-shot request.

    Naming Consistency5/5

    All tool names consistently use snake_case with a clear get_ or list_ verb prefix followed by a specific resource. The convention is predictable throughout the set.

    Tool Count5/5

    Nine tools is well-scoped for a machine-monitoring server. Each tool maps to a specific dashboard area or data view without excessive granularity.

    Completeness4/5

    The read-only monitoring surface is broad: machines, status, settings, firmware, schedules, counters, recent shots, and trends are all covered. Missing control actions such as changing schedules or firmware updates are noticeable, but may be intentionally outside the API's scope.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      A Machine Context Protocol server that enables Claude AI to interact with tools through a structured communication interface, following standard MCP patterns with server initialization and stdio transport.
      2,530 npm
      2
      MIT
    • F
      license
      Not graded
      quality
      A
      maintenance
      Exposes a governed suite of 38 tools over a single OAuth-gated MCP endpoint, letting web AI chat clients such as Claude, ChatGPT, and Grok operate on a local machine — reading and editing files, running whitelisted shell commands, searching, querying SQLite, driving browsers and DevTools, managing background jobs, ports, and clipboard. Local IDEs and Postman can also connect directly over localhost for offline, low-latency access.
      -
    • A
      license
      Not graded
      quality
      A
      maintenance
      Enables MCP-compatible clients to manage, run, and interact with local or cloud LLM models through exposed tools for listing, pulling, generating, chatting, embeddings, and web search/fetch. Supports Claude Desktop, Cline, and hybrid local/cloud setups.
      812 npm
      AGPL 3.0