Sugar gives AI coding agents a persistent, cross-project memory via MCP.
Most agents start every session knowing nothing about your codebase, your preferences, or decisions you made last week. Sugar gives them a memory that survives sessions, survives switching tools, and survives switching projects.
It stores typed memories: decisions, preferences, file context, error patterns, research notes, outcomes, and guidelines. When your agent starts a session, it can pull relevant context from ~/.sugar/memory.db - shared across all your projects. Switch from Claude Code to Goose mid-week? Same memory.
Tech: Python, SQLite + FTS5, sentence-transformers for vector search with FTS5/LIKE fallback. No server. No Docker. No cloud account.
pip install sugarai
Works with Claude Code, OpenCode, and Goose today (anything that speaks MCP).
Most agents start every session knowing nothing about your codebase, your preferences, or decisions you made last week. Sugar gives them a memory that survives sessions, survives switching tools, and survives switching projects.
It stores typed memories: decisions, preferences, file context, error patterns, research notes, outcomes, and guidelines. When your agent starts a session, it can pull relevant context from ~/.sugar/memory.db - shared across all your projects. Switch from Claude Code to Goose mid-week? Same memory.
Tech: Python, SQLite + FTS5, sentence-transformers for vector search with FTS5/LIKE fallback. No server. No Docker. No cloud account.
Works with Claude Code, OpenCode, and Goose today (anything that speaks MCP).
AGPL-3.0. GitHub: https://github.com/roboticforce/sugar
Happy to answer questions about what works, what doesn't, and where it's rough around the edges.