Detailed Introduction
grov is a collective memory tool for engineering teams that captures the reasoning, decisions, and context produced when a developer interacts with an AI agent (for example, Claude Code), and syncs those memories to a team dashboard or local storage. Using a local proxy with optional team sync, grov converts one developer’s discoveries into injectible context so subsequent sessions avoid repeated exploration and token cost.
Main Features
- Capture structured reasoning traces and key decisions from agent sessions.
- Local-first storage in
~/.grov/memory.db(SQLite) with optional team sync. - CLI and local proxy with extended cache, auto-compaction, and anti-drift detection.
- Hybrid search (semantic + lexical) and automatic injection of team context into new sessions.
Use Cases
Ideal for teams that want agents to share knowledge across sessions: reducing redundant exploration, surfacing verified project rationale during code review or design discussions, and onboarding new members or assistants with accurate project context. Particularly useful for workflows built around Claude Code.
Technical Features
grov combines a local proxy, LLM-based extraction, SQLite storage, and compression strategies to preserve essential reasoning while keeping context windows manageable. It exposes a CLI, dashboard, and adapters for integrating with various agent runtimes and team environments.