This project provides a "free build" of Anthropic's Claude Code CLI, stripping telemetry and security guardrails while enabling experimental features. It targets developers and power users seeking an unconstrained, locally-controlled AI coding assistant, offering enhanced functionality and privacy by eliminating external reporting and prompt-level restrictions.

How It Works

This fork is based on a snapshot of Anthropic's Claude Code CLI, modified to remove all outbound telemetry (OpenTelemetry, Sentry, GrowthBook analytics) and Anthropic's injected security guardrails. It unlocks 54 of the 88 available experimental features that compile cleanly, offering advanced capabilities like remote multi-agent planning and IDE bridging. The project utilizes the Bun runtime and TypeScript, supporting multiple LLM providers (Anthropic, OpenAI, AWS Bedrock, Google Vertex AI, Anthropic Foundry) configurable via environment variables.

Quick Start & Requirements

• Install: Execute curl -fsSL https://raw.githubusercontent.com/paoloanzn/free-code/main/install.sh | bash. This script checks system requirements, installs Bun if necessary, clones the repository, builds the CLI with experimental features enabled, and adds it to the system's PATH.
• Prerequisites: Bun version 1.3.11 or higher, macOS or Linux operating system (Windows via WSL). An API key or OAuth login for the chosen model provider is required.
• Build: Clone the repository (git clone), navigate into the directory (cd free-code), and run bun build followed by ./cli.
• Links: Install script: https://raw.githubusercontent.com/paoloanzn/free-code/main/install.sh. Feature audit: FEATURES.md.

Highlighted Details

• Telemetry Stripped: All outbound telemetry endpoints are dead-code-eliminated or stubbed, ensuring no usage analytics or error reporting.
• Guardrails Removed: System-level instructions and hardcoded refusal patterns injected by Anthropic are stripped, relying only on the model's inherent safety training.
• Experimental Features Unlocked: Includes features like ULTRAPLAN for remote multi-agent planning, ULTRATHINK for enhanced reasoning, VOICE_MODE for voice input, and BRIDGE_MODE for IDE remote control.
• Multi-Provider Support: Seamlessly switch between Anthropic, OpenAI Codex, AWS Bedrock, Google Vertex AI, and Anthropic Foundry via environment variables.

Maintenance & Community

The README notes that the "Parent repo is migrating ownership, until the operation is done this repo is blocked," indicating potential instability or unavailability. No specific community links (Discord, Slack) or details on active maintenance or contributors are provided.

Licensing & Compatibility

The original Claude Code source is proprietary to Anthropic. This fork's license is not explicitly stated, but it is derived from source code exposed via npm distribution. Users are advised to "Use at your own discretion," implying potential legal ambiguities and restrictions for commercial use or linking within closed-source projects.

Limitations & Caveats

The project is currently "blocked" due to parent repository ownership migration, posing a significant adoption risk. It requires macOS or Linux environments. Of the 88 experimental features, 34 are noted as broken, and the licensing status is unclear, derived from proprietary code without an explicit open-source license for the fork.