Version:
← All Docs

Goose Integration

Sugar integrates with Goose (Block's open-source AI agent) via MCP, providing task management and memory features directly in your Goose sessions.

Overview

The Goose integration enables:

  • Task management - Create, list, and manage Sugar tasks from Goose
  • Memory access - Store and recall decisions, preferences, and patterns
  • Project context - Get organized summaries of your project knowledge

Quick Start

Option 1: Goose CLI (Recommended)

goose configure
# Select "Add Extension" → "Command-line Extension"
# Name: sugar
# Command: npx -y sugarai-mcp

Option 2: Direct Config Edit

Add to ~/.config/goose/config.yaml:

extensions:
  sugar:
    command: npx -y sugarai-mcp

After setup, restart Goose to load the Sugar extension.

Prerequisites

  • Goose installed
  • Node.js (for npx)
  • Sugar initialized in your project (sugar init)

Available MCP Tools

Once configured, Goose has access to these Sugar tools:

Task Management

Tool Description
createTask Add a new task to Sugar's queue
listTasks List tasks with optional filtering
viewTask Get detailed information about a task
updateTask Modify task priority, status, or metadata
removeTask Delete a task from the queue
getStatus Get Sugar system status and queue stats

Memory System

Tool Description
search_memory Semantic search over stored memories
store_learning Store new decisions, preferences, or patterns
get_project_context Get organized project summary
recall Get formatted markdown context on a topic
list_recent_memories List memories with optional type filter

Usage Examples

Once Sugar is configured in Goose, you can interact naturally:

Task Management

"Add a task to fix the login timeout bug with high priority"
"Show me all pending bug fixes"
"What's the status of the Sugar queue?"

Memory

"Remember that we use JWT with RS256 for authentication"
"What decisions have we made about the database?"
"Store this as an error pattern: connection timeouts need retry logic"

Alternative: Python MCP Server

If you prefer not to use npx, Sugar also provides a built-in Python MCP server:

# Start the HTTP-based MCP server
sugar mcp serve --port 8080

Then configure Goose to connect to http://localhost:8080.

Troubleshooting

"Extension not found" or tools not available

  1. Verify Node.js is installed: node --version
  2. Test the MCP server manually: npx -y sugarai-mcp
  3. Restart Goose after configuration changes

"Not a Sugar project"

Sugar requires initialization in your project directory:

cd /path/to/your/project
sugar init

Tools work but no memories/tasks

Ensure you're running Goose from a directory with a .sugar/ folder, or that SUGAR_PROJECT_ROOT is set.

Environment Variables

Variable Description Default
SUGAR_PROJECT_ROOT Override project directory Current directory
SUGAR_LOG_LEVEL Logging verbosity INFO

Learn More