CodeFreedom
One CLI for every code agent.
Switch LLM providers, isolate environments, and manage everything from ~/.codefreedom.
pip install codefreedom
What Problem Does This Solve¶
You have code agents (Claude Code, MiMo Code, OpenCode, etc.). You want to switch between AI models (DeepSeek, GPT, Claude) without reconfiguring everything. You want isolated environments. You want one place for all your settings.
CodeFreedom gives you that.
What You Get¶
- Switch models instantly
DeepSeek for drafting, GPT for reasoning, free models for testing. Same command, different profile. No code changes.
- Isolated sandboxes
Every session runs in a fresh Docker container. CUDA, ROCm, or plain Ubuntu. Clean slate every time.
- Self-hosted proxy
One local endpoint (localhost:4000) routes to any LLM provider. Add API keys, switch backends, track spend.
- Browser and API tools
Headless Chrome, web search, GitHub API — all as Docker containers your code agent can use.
How It Works¶
- You run
cf run agent claude-code - CodeFreedom sets up environment variables from your profile
- Claude Code talks to your local proxy at
localhost:4000 - The proxy routes the request to whichever AI model you configured
Change the model? Edit one line in your profile. Done.
Quick Start¶
Four commands, five minutes:
pip install codefreedom # Install
cf setup init # Set up config via recipe
cf run proxy start # Start the proxy
cf run agent claude-code # Launch Claude Code
Short aliases¶
All commands have short aliases for faster workflows: