Free Code is a terminal-native AI coding agent that edits files, runs shell commands, and answers questions through natural language conversations directly in the command line. It operates as an interactive loop in the shell, providing direct filesystem manipulation without intermediate abstraction layers or sandboxing.
The project distinguishes itself through a compile-time feature flag system that activates or deactivates 54 experimental capabilities, including voice input, multi-agent planning, and memory extraction. It also routes all AI requests to one of five supported API backends by setting a single environment variable, and includes a multi-agent browser-based planning mode that runs autonomous research and planning tasks for 10-30 minutes. The build process strips all telemetry collection and security-prompt guardrails from the original codebase.
The documentation covers installation and configuration for the terminal-native agent loop, the environment-variable-based provider routing mechanism, and the experimental feature flag system.