OpenWolf provides middleware for Claude Code, addressing its lack of project awareness and memory. It enhances Claude's efficiency by offering a project index, learning memory, and token tracking, aiming to reduce token consumption and prevent repeated errors without altering user workflows.

How It Works

OpenWolf integrates via 6 Node.js hook scripts, creating a .wolf/ directory. This directory houses anatomy.md (project file map with token estimates), cerebrum.md (learning memory for corrections and preferences), memory.md (action log), buglog.json (bug fix repository), and token-ledger.json (token usage tracking). This "second brain" allows Claude to infer file content, avoid redundant reads, learn from mistakes, and manage token usage, leading to significant efficiency gains.

Quick Start & Requirements

• Install: npm install -g openwolf
• Initialize: Navigate to your project directory and run openwolf init.
• Prerequisites: Node.js 20+ and the Claude Code CLI.
• Optional: PM2 for background task persistence, puppeteer-core for Design QC. Setup is minimal.

Highlighted Details

• Token Efficiency: Claims average 65.8% token reduction across numerous sessions, with ~71% of repeated file reads blocked.
• Design QC: Enables full-page application screenshot capture for Claude-driven design evaluation.
• Framework Reframe: Offers curated knowledge bases and migration prompts for 12 UI frameworks to assist Claude in selection and implementation.
• Zero Workflow Change: Operates via background hooks, making its intelligence invisible to the user.

Maintenance & Community

This project is at v1.0.4. The README lists Farhan Palathinkal Afsal (Cytostack) as the author. No community channels (Discord, Slack) or explicit roadmap details are provided.

Licensing & Compatibility

• License: AGPL-3.0.
• Compatibility: The AGPL-3.0 license is a strong copyleft license. Its use may require derivative works to be open-sourced under the same license, potentially impacting integration with closed-source commercial products.

Limitations & Caveats

• Hook Dependency: Relies on Claude Code's hook feature, falling back to CLAUDE.md if hooks fail.
• Token Estimation: Token counts are estimations (~15% accuracy), not precise API figures.
• Learning Compliance: cerebrum.md effectiveness depends on Claude adhering to update instructions (~85-90% compliance).
• Stability: As v1.0.4, potential for bugs or breaking changes exists.