Skip to main content
Glama

Characters live in a pixel-art room, perform activities based on agent state, show emotions, and have creature companions. View them at life.cubeworld.co.za, on the desktop companion, or embedded in your own app via <iframe>.

Packages

This repo contains three packages:

Package

Install

Description

@cubelife/sdk

npm install @cubelife/sdk

Node.js SDK for state reporting

cubelife

npm install -g cubelife

CLI + MCP server for AI coding tools

cubelife

pip install cubelife

Python SDK for state reporting

Related MCP server: The Agents MCP Server

Quick start

MCP server (for Claude Code, Cursor, Copilot, Windsurf, Cline, Codex)

npx cubelife tutorial           # create account + agent
npx cubelife setup claude-code  # auto-configure MCP

Your agent gets four tools: cubelife_report, cubelife_complete, cubelife_error, cubelife_status.

Node SDK

npm install @cubelife/sdk
import { CubeLifeClient } from '@cubelife/sdk';

const client = new CubeLifeClient({ apiKey: process.env.CUBELIFE_API_KEY });
await client.report('coding', { detail: 'Building the login page', progress: 0.5 });
await client.report('complete', { detail: 'Login page shipped' });

Python SDK

pip install cubelife
from cubelife import CubeLifeClient

client = CubeLifeClient("your-agent-key")
client.report("coding", detail="Building the login page", progress=0.5)
client.report("complete", detail="Login page shipped")

Creature companions

Every agent gets a creature companion (fox, owl, cat, dragon, wolf, phoenix, or slime) that reacts alongside the character.

MCP server

The CLI includes an MCP server for AI coding tools. Auto-configure for your tool:

npx cubelife setup claude-code    # Claude Code
npx cubelife setup cursor         # Cursor
npx cubelife setup windsurf       # Windsurf
npx cubelife setup copilot        # GitHub Copilot
npx cubelife setup cline          # Cline
npx cubelife setup codex          # Codex

Or manually: claude mcp add cubelife -- npx cubelife mcp

Licence

MIT (source code). Pixel art sprite assets are proprietary (see SPRITE-LICENSE).

Available Tools

4 tools
cubelife_completeA

Mark the current task as complete. Triggers a celebration animation on the character.

Inputs: detail.

cubelife_errorB

Report that an error occurred. Shows frustration on the character.

Inputs: detail.

cubelife_reportA

Report the AI agent's current work state to CubeLife. Call this when starting a new task or changing activity.

Inputs: state (required), detail, progress, sentiment.

cubelife_statusA

Get the agent's current state.

Inputs: no named parameters.

Tool Schema Changelog

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

Loading tool schema changelog…

    TDQS

    A3.9/5.0

    Scored across 4 tools

    Disambiguation5/5

    Each tool has a distinct purpose: marking completion, reporting errors, updating work state, and retrieving status. No overlap in functionality.

    Naming Consistency4/5

    All tools follow the 'cubelife_verb' pattern with underscore, though 'error' and 'status' are nouns used as verbs, causing slight inconsistency.

    Tool Count5/5

    Four tools are appropriate for managing an agent's lifecycle and state reporting, covering key actions without being excessive or insufficient.

    Completeness4/5

    The surface covers starting/changing tasks, completion, error reporting, and status queries. Missing a cancel or pause operation, but core functionality is present.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers