oobo

Product Introduction

1. Definition: oobo is an open-source, Rust-based Git decorator and metadata enrichment engine designed for the era of AI-augmented software development. It functions as a transparent wrapper for the Git CLI, capturing and linking high-fidelity AI session data—such as conversation transcripts, token usage, and model attribution—directly to version control commits.

2. Core Value Proposition: oobo exists to solve the "Context Gap" in modern programming. While traditional Git tracks what changed (the diff), it fails to capture the why and how behind AI-generated code. oobo enriches repositories with AI Contextual Persistence, allowing both human developers and autonomous AI agents to access the full decision-making history, chat logs, and resource costs associated with every line of code.

Main Features

1. Transparent Git Interception: oobo operates as a pass-through decorator for the Git command line. When a user executes a "write" operation (like oobo commit or oobo push), the tool intercepts the call, executes the standard Git logic, and simultaneously scans local AI tool directories (e.g., Cursor, Claude Code, Zed). It uses high-performance Rust internals to ensure zero overhead on standard "read" operations like status or diff.

2. The Anchor Metadata System: This is oobo's core primitive. An "Anchor" is a structured metadata object that extends a standard Git commit. It includes:

   • Session Linking: UUIDs of the AI chat sessions that contributed to the change.
   • Token Analytics: Exact counts of input and output tokens used during the development of that specific diff.
   • Code Attribution: Statistical breakdown of AI-generated vs. human-authored lines.
   • Model Provenance: Data on which specific LLM (e.g., Claude 3.5 Sonnet, GPT-4o) was utilized. Anchors are stored in a local SQLite database and synchronized via a specialized Git orphan branch (oobo/anchors/v1), ensuring context travels with the repository without polluting the main code history.

3. Agent-Ready JSON Interface: Built with "Agents-as-Users" in mind, oobo features a global --agent flag. This flag transforms all CLI output into structured, machine-readable JSON. Additionally, oobo automatically manages Agent Skill Files (SKILL.md), which are placed in standard directories searched by AI coding tools. This allows agents to self-install oobo and use it to document their own work programmatically.

4. Integrated Session TUI & Analytics: Beyond tracking, oobo provides a Terminal User Interface (TUI) for browsing cross-tool AI sessions. Users can search through transcripts from different IDEs (Cursor, Windsurf) and CLI tools (Aider, Claude Code) in one unified view. The oobo stats command provides project-wide insights into productivity metrics, model costs, and AI contribution ratios.

Problems Solved

1. Pain Point: Ephemeral AI Context: AI coding tools often generate deep reasoning and complex debugging sessions that vanish when a chat tab is closed or a history is cleared. oobo solves this by anchoring those conversations to the permanent Git history.

2. Target Audience:

   • AI-First Software Engineers: Developers using Cursor, Zed, or Claude Code who want to maintain a searchable history of their AI interactions.
   • Engineering Leaders: Managers needing to audit AI usage, monitor token costs, and understand the "AI-to-Human" code ratio across teams.
   • Autonomous AI Agents: Agents that need to "read the room" by understanding the history of previous agentic interventions in a codebase.
   • Open Source Maintainers: Who need to verify the provenance and "human-in-the-loop" verification of AI-submitted Pull Requests.

3. Use Cases:

   • Post-Mortem Analysis: Using oobo sessions to review the exact 20,000-token conversation that led to a specific architectural decision during a high-pressure incident.
   • Token Budgeting: Tracking real-world costs of different AI models across various projects to optimize IDE and API spend.
   • Seamless Onboarding: New team members reading the AI "thought process" behind complex legacy modules instead of just reading the raw code.

Unique Advantages

1. Zero-Configuration Workflow: By aliasing git=oobo, developers can use their existing Git habits. There are no new commands to learn for basic version control; the AI enrichment happens automatically in the background.

2. Local-First Privacy & Security: Unlike cloud-based AI analytics platforms, oobo keeps all transcripts and session data on the local machine by default. It includes built-in secret redaction using gitleaks patterns to ensure API keys or passwords never accidentally enter the metadata anchors.

3. Decentralized Synchronization: Because oobo uses an orphan Git branch to store anchors, the AI context is synchronized naturally through git push and git pull. This eliminates the need for a centralized "oobo server," making it highly resilient and compatible with existing CI/CD pipelines.

Frequently Asked Questions (FAQ)

1. Does oobo slow down my Git commands? No. oobo is written in Rust for maximum efficiency. It uses a non-blocking architecture for read operations, passing them directly to the Git binary. Metadata collection only occurs during write operations (commits/merges), adding only a few milliseconds to the process.

2. Is my AI chat history shared with other developers? By default, session transcripts stay local to your machine. Only the "Anchor" metadata (session IDs, token counts, attribution) is synced via the Git orphan branch. You can explicitly enable "Transparency Mode" if you wish to share full transcripts with your team for collaborative debugging.

3. How does oobo identify which AI session belongs to which commit? oobo uses a combination of time-window correlation and "Agent Lifecycle Hooks." For supported tools like Claude Code and Gemini CLI, oobo installs hooks that notify it exactly when a session starts and ends, ensuring 100% accuracy in linking conversations to code changes.