sugar

Overview

Sugar is a sophisticated memory system for AI coding agents that combines persistent storage with autonomous task execution. It operates through MCP servers that expose tools for searching and storing memories, as well as executing tasks autonomously. The system uses a dual-database approach (project-specific and global) with semantic search capabilities to provide relevant context to AI agents across sessions. This enables AI assistants to remember project-specific decisions, preferences, and error patterns while also accessing cross-project guidelines and best practices.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• search_memory

  Search both memory stores, returns results with scope labels

  Try:Search for authentication patterns in the projectTry:Find guidelines for database queries
• store_learning

  Save a memory (pass scope: 'global' for cross-project)

  Try:Remember to always validate user inputTry:Store this as a global guideline
• recall

  Get formatted markdown context for a topic

  Try:Recall information about async patternsTry:What did we decide about error handling?
• get_project_context

  Full project summary including global guidelines

  Try:Show me the project contextTry:Get an overview of this project's decisions
• list_recent_memories

  Browse recent memories by type

  Try:List recent decisionsTry:Show recent error patterns

Comparable tools

Installation

# Recommended installation with pipx
pipx install sugarai

# With semantic search (recommended)
pipx install 'sugarai[memory]'

# With GitHub integration
pipx install 'sugarai[github]'

# Setup for Claude Code
claude mcp add sugar -- sugar mcp memory
claude mcp add sugar-tasks -- sugar mcp tasks

FAQ

What memory types does Sugar support?

Sugar supports seven memory types: decision, preference, file_context, error_pattern, research, outcome, and guideline. Each has different TTL and retrieval behavior.

How does Sugar handle both project-specific and global memory?

Sugar uses two SQLite databases: one project-specific (.sugar/memory.db) and one global (~/.sugar/memory.db). Search is project-first but always includes global guidelines.

On Hacker News

Recent discussion from the developer community.

• Show HN: Sugar – Cross-project memory for AI coding agents via MCP▲ 1
  Story by cdnsteve · 2026-03-17
• Show HN: Sugar – A task queue that lets AI coding agents work autonomously▲ 1
  Story by cdnsteve · 2026-02-27

Compare sugar with