🍰 Sugar

The autonomous layer for AI coding agents.

Sugar manages your task queue, runs 24/7, and ships working code while you focus on what matters.

What It Does

Sugar adds autonomy and persistence to your AI coding workflow. Instead of one-off interactions:

• Continuous execution - Runs 24/7, working through your task queue
• Agent-agnostic - Works with Claude Code, OpenCode, Aider, or any AI CLI
• Persistent memory - Remember decisions, preferences, and patterns across sessions
• Delegate and forget - Hand off tasks from any session
• Builds features - Takes specs, implements, tests, commits working code
• Fixes bugs - Reads error logs, investigates, implements fixes
• GitHub integration - Creates PRs, updates issues, tracks progress

You plan the work. Sugar executes it.

Works with: Claude Code | OpenCode | Aider | Goose | Any CLI-based AI agent

Native AI Agent Integrations

Sugar has first-class integrations with leading AI coding agents:

Agent	Integration	Features
Claude Code	MCP Server	Memory access, task delegation, context injection
OpenCode	Plugin + HTTP API	Bidirectional communication, notifications, memory injection
Goose	MCP Server	Task management, memory access

Claude Code - Full memory system access via MCP:

claude mcp add sugar -- sugar mcp memory

OpenCode - One-command setup with MCP integration:

sugar opencode setup  # Automatically configures OpenCode
# Then restart OpenCode

Both integrations support automatic memory injection - Sugar injects relevant context (decisions, preferences, error patterns) into your AI sessions automatically.

Install

Recommended: pipx (install once, use everywhere)

pipx install sugarai

This gives you a global sugar command that works in any project. Each project gets its own isolated config and database in a .sugar/ folder.

Upgrade / Uninstall:

pipx upgrade sugarai    # Upgrade to latest version
pipx uninstall sugarai  # Remove completely

Other installation methods

pip (requires venv activation each session)

pip install sugarai

uv (fast alternative to pip)

uv pip install sugarai

With GitHub integration:

pipx install 'sugarai[github]'

With memory system (semantic search):

pipx install 'sugarai[memory]'

All features:

pipx install 'sugarai[all]'

Quick Start

Sugar is project-local - each project has its own isolated task queue and config.

# Navigate to your project
cd ~/dev/my-app

# Initialize Sugar (creates .sugar/ folder)
sugar init

# This creates:
# - .sugar/sugar.db     (task queue database)
# - .sugar/config.yaml  (project settings)
# - .sugar/prompts/     (custom prompts)

# Add tasks to the queue
sugar add "Fix authentication timeout" --type bug_fix --urgent
sugar add "Add user profile settings" --type feature

# Start the autonomous loop
sugar run

Sugar will:

1. Pick up tasks from the queue
2. Execute them using your configured AI agent
3. Run tests and verify changes
4. Commit working code
5. Move to the next task

It keeps going until the queue is empty (or you stop it).

Memory System

Sugar remembers what matters across sessions. No more re-explaining decisions or rediscovering patterns.

Saves tokens: Memories are stored as compressed summaries (~90% smaller) and retrieved only when relevant. Real projects see ~89% token reduction per session - that's ~$32 saved over 500 sessions.

# Store knowledge
sugar remember "Always use async/await, never callbacks" --type preference
sugar remember "JWT tokens use RS256, expire in 15 min" --type decision

# Search memories
sugar recall "authentication"

# Claude Code integration - give Claude access to your project memory
claude mcp add sugar -- sugar mcp memory

# See your token savings
python examples/token_savings_demo.py

Memory types: decision, preference, file_context, error_pattern, research, outcome

Full docs: Memory System Guide

Delegate from Claude Code:

/sugar-task "Fix login timeout" --type bug_fix --urgent

Sugar picks it up and works on it while you keep coding.

How It Works: Project-Local Architecture

Global Installation (pipx)
└── sugar CLI (~/.local/bin/sugar)

Project A                          Project B
~/dev/frontend-app/                ~/dev/backend-api/
├── .sugar/                        ├── .sugar/
│   ├── sugar.db                   │   ├── sugar.db
│   ├── config.yaml                │   ├── config.yaml
│   └── prompts/                   │   └── prompts/
├── src/                           ├── main.py
└── tests/                         └── requirements.txt

Running `sugar` uses the .sugar/ folder in your current directory

One global CLI, many isolated projects. Like git - one installation, per-project repositories.

FAQ

Do I need to install Sugar in every project?

No! Install Sugar once with pipx install sugarai and use it everywhere.

The sugar command is globally available, but it reads configuration from the .sugar/ folder in your current directory:

• Global CLI access: Run sugar from anywhere without venv activation
• Project-local state: Each project's tasks and config stay isolated
• No conflicts: Work on multiple projects simultaneously

Can I run Sugar on multiple projects at the same time?

Yes! Each project has its own isolated database.

# Terminal 1
cd ~/dev/frontend-app
sugar run

# Terminal 2 (simultaneously)
cd ~/dev/backend-api
sugar run

The two Sugar instances won't interfere with each other.

What happens if I run sugar outside a project folder?

Sugar will show a friendly error:

❌ Not a Sugar project

   Could not find: .sugar/config.yaml

   Run 'sugar init' to initialize Sugar in this directory.

Why pipx over pip?

Installation	Global access?	Requires venv?
pip install sugarai	Only in active venv	Yes
pipx install sugarai	Yes, always	No

With pipx, Sugar's dependencies don't conflict with your project's dependencies.

Should I commit .sugar/ to git?

Recommended .gitignore:

.sugar/sugar.db       # Task queue is personal
.sugar/sugar.log      # Logs contain local paths
.sugar/*.db-*         # SQLite temp files

DO commit: .sugar/config.yaml and .sugar/prompts/ to share settings with your team.

Features

Native AI Agent Integrations (New in 3.5)

• Claude Code - MCP server for memory access and task delegation
• OpenCode - Full bidirectional integration with notifications and context injection
• Automatic memory injection into AI sessions
• Real-time task lifecycle notifications

Memory System (New in 3.5)

• Persistent semantic memory across sessions
• Remember decisions, preferences, error patterns
• Claude Code & OpenCode integration via MCP server
• Semantic search with sugar recall

Task Management

• Rich task context with priorities and metadata
• Custom task types for your workflow
• Queue management and filtering

Task Orchestration

• Auto-decomposes complex features into subtasks
• 4-stage workflow: Research → Planning → Implementation → Review
• Specialist agent routing (frontend, backend, QA, security, DevOps)
• Parallel execution with dependency management

Autonomous Execution

• Specialized task agents (UX, backend, QA, security, DevOps)
• Automatic retries on failures
• Quality checks and testing

GitHub Integration

• Reads issues, creates PRs
• Updates issue status automatically
• Commits with proper messages

Ralph Wiggum Integration

• Iterative execution for complex tasks
• Self-correcting loops until tests pass
• Prevents single-shot failures

Full docs: Memory System | OpenCode Integration | Ralph Wiggum

Configuration

.sugar/config.yaml is auto-generated on sugar init. Key settings:

sugar:
  dry_run: false              # Set to true for testing
  loop_interval: 300          # 5 minutes between cycles
  max_concurrent_work: 3      # Parallel task execution

claude:
  enable_agents: true         # Use specialized Claude agents

discovery:
  github:
    enabled: true
    repo: "user/repository"
  error_logs:
    enabled: true
    paths: ["logs/errors/"]

Integrations

Claude Code Plugin

Sugar has native Claude Code integration. Delegate work directly from your Claude sessions.

/plugin install roboticforce/sugar

Inside a Claude Code session:

You: "I'm working on auth but need to fix these test failures.
     Can you handle the tests while I finish?"

Claude: "I'll create a Sugar task for the test fixes."

/sugar-task "Fix authentication test failures" --type test --urgent

Available Slash Commands:

• /sugar-task - Create tasks with rich context
• /sugar-status - Check queue and progress
• /sugar-run - Start autonomous mode

OpenCode Integration

Sugar has native MCP integration with OpenCode:

Features:

• MCP servers for task management and memory access
• Automatic memory injection into OpenCode sessions
• Context-aware memory retrieval based on current work

Quick Setup:

# One-command setup - adds Sugar MCP servers to OpenCode config
sugar opencode setup

# Restart OpenCode to load the new servers
# Then verify:
sugar opencode status

The setup command automatically:

• Finds your OpenCode config file
• Adds sugar-tasks and sugar-memory MCP servers
• Preserves your existing configuration

Options:

sugar opencode setup --dry-run  # Preview changes without applying
sugar opencode setup --yes      # Non-interactive mode
sugar opencode setup --no-memory  # Only add task server

MCP Server Integration

Sugar provides MCP servers for Goose, Claude Code, Claude Desktop, and other MCP clients.

Using with Claude Code (Memory):

# Add Sugar memory to Claude Code
claude mcp add sugar -- sugar mcp memory

Or add to ~/.claude.json:

{
  "mcpServers": {
    "sugar": {
      "type": "stdio",
      "command": "sugar",
      "args": ["mcp", "memory"]
    }
  }
}

This gives Claude Code access to your project's memory - decisions, preferences, error patterns, and more.

Using with Goose:

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

Using with Claude Desktop:

{
  "mcpServers": {
    "sugar": {
      "command": "npx",
      "args": ["-y", "sugarai-mcp"],
      "env": {
        "SUGAR_PROJECT_ROOT": "/path/to/your/project"
      }
    }
  }
}

Memory System

Sugar's memory system provides persistent context across sessions:

# Store memories
sugar remember "Always use async/await, never callbacks" --type preference
sugar remember "Auth tokens expire after 15 minutes" --type research --ttl 90d

# Search memories
sugar recall "how do we handle authentication"
sugar recall "error patterns" --type error_pattern

# List and manage
sugar memories --type decision --since 7d
sugar forget abc123 --force
sugar memory-stats

# Export for Claude Code SessionStart hook
sugar export-context

Memory types: decision, preference, file_context, error_pattern, research, outcome

Full docs: Memory System Guide

Advanced Usage

Task Orchestration

sugar add "Add OAuth authentication" --type feature --orchestrate

# Sugar will:
# 1. RESEARCH - Search best practices, analyze codebase
# 2. PLAN - Create implementation plan with subtasks
# 3. IMPLEMENT - Route subtasks to specialists in parallel
# 4. REVIEW - Code review and test verification

sugar orchestrate <task_id> --stages

Ralph Wiggum (iterative execution)

sugar add "Implement rate limiting" --ralph --max-iterations 10
# Iterates until tests pass, not just until code is written

Custom Task Types

sugar task-type add deployment --name "Deployment" --emoji "🚀"
sugar add "Deploy to staging" --type deployment

Complex Tasks with Context

sugar add "User Dashboard" --json --description '{
  "priority": 5,
  "context": "Complete dashboard redesign",
  "agent_assignments": {
    "frontend_developer": "Implementation",
    "qa_test_engineer": "Testing"
  }
}'

Troubleshooting

Sugar not finding Claude CLI?

# .sugar/config.yaml
claude:
  command: "/full/path/to/claude"

Tasks not executing?

cat .sugar/config.yaml | grep dry_run  # Check dry_run is false
tail -f .sugar/sugar.log                # Monitor logs
sugar run --once                        # Test single cycle

More help:

• Troubleshooting Guide
• GitHub Issues

Documentation

• Quick Start
• CLI Reference
• Memory System (New)
• Task Orchestration
• Ralph Wiggum
• GitHub Integration
• Configuration Guide

Requirements

• Python 3.11+
• An AI coding agent CLI:
  • Claude Code (default)
  • OpenCode
  • Aider
  • Or any CLI-based AI coding tool

Contributing

Contributions welcome! See CONTRIBUTING.md.

git clone https://github.com/roboticforce/sugar.git
cd sugar
uv pip install -e ".[dev,test,github]"
pytest tests/ -v

License

Dual License: AGPL-3.0 + Commercial

• Open Source (AGPL-3.0): Free for open source and personal use
• Commercial License: For proprietary use - sugar.roboticforce.io/licensing

---

Sugar v3.5 - The autonomous layer for AI coding agents

Now with native Claude Code and OpenCode integrations for seamless AI agent collaboration.

⚠️ Sugar is provided "AS IS" without warranty. Review all AI-generated code before use.