Skip to content

CodeFreedom

One CLI for every code agent. Switch LLM providers, isolate environments, and manage everything from ~/.codefreedom. 

uv tool install codefreedom
Get Started GitHub

The Problem

AI code agents are powerful, but the ecosystem around them is fragmented. Each agent carries its own configuration, model preferences, and runtime dependencies. There is no portable layer that lets you move between agents, models, or providers without starting over.

That fragmentation creates three compounding risks:

  • Vendor lock-in -- choosing a provider today can constrain your choices tomorrow. As the model landscape evolves, being tied to one ecosystem means missing better capabilities, pricing, or reliability from competitors.
  • Unmanaged cost -- without visibility across providers, there is no way to route work to the most cost-effective model. Token spend grows unchecked because switching providers is a manual, error-prone process.
  • Complexity as a barrier -- setting up proxies, sandboxes, and provider integrations demands infrastructure expertise. Developers who should be building products spend time plumbing tooling, or avoid the tools altogether.

What You Get

  • Switch models instantly — DeepSeek for drafting, GPT for reasoning, free models for testing. Same command, different profile.
  • Isolated sandboxes — Every session runs in a fresh Docker container. Clean slate every time.
  • Self-hosted proxy — One local endpoint (localhost:4000) routes to any LLM provider.
  • Browser and API tools — Headless Chrome, web search, GitHub API — all as Docker containers.

How It Works

You → cf run agent claude-code → LiteLLM Proxy → Your chosen AI model
  1. You run cf run agent claude-code
  2. CodeFreedom sets up environment variables from your profile
  3. Claude Code talks to your local proxy at localhost:4000
  4. The proxy routes the request to whichever AI model you configured

Change the model? Edit one line in your profile. Done.

Quick Start

uv tool install codefreedom                              # Install
cf s i -pa costeffective-coding                          # Set up recipe
cf r px start                                            # Start the proxy
cf r ag cc                                               # Launch Claude Code

Get started step by step →