Claude Zen — Development skill for Claude Code
Run Claude Code on free models through OpenCode Zen.
How to install Claude Zen
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open Itsme23476/claude-zen and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Claude Zen does
Run Claude Code on free models through OpenCode Zen. Local Anthropic-protocol bridge + one-command install.
Alternatives in Development
- Magnitude — Open source inference server your agent sets up for you 1.5k ★
- Claudecode — Open Source ClaudeCode Leaked 1k ★
- Hive — Browser-native hive-mind for CLI coding agents — Claude Code, Codex, Gemini, and OpenCode collaborate as real 473 ★
README
claude-zen
Run **Claude Code** on free models through [OpenCode Zen](https://opencode.ai) — a small local proxy that translates Zen's OpenAI-style API into the Anthropic protocol Claude Code speaks.
Default model is `deepseek-v4-flash-free`. Cost: $0.
Unofficial community tool. Not affiliated with or endorsed by Anthropic or OpenCode. It relies on a third party's free tier, which can change or disappear at any time.
**Requirements:** macOS or Linux, Node 18+, Python 3, and Claude Code installed. Windows works under WSL only.
Install
git clone https://github.com/Itsme23476/claude-zen
cd claude-zen
./install.sh # asks for your Zen API key
./verify.sh # proves it works
Get a free key at [opencode.ai](https://opencode.ai). It looks like `sk-` plus 61 characters.
Then:
claude-zen
That's it. The proxy auto-starts if it isn't running.
Usage
claude-zen # interactive, default free model
claude-zen -m longcat-2.0-free # pick a specific model
claude-zen -p "explain this bug" # one-shot, no TUI
claude-zen --status # proxy state + live free-model list
Anything after those flags is passed straight through to `claude`, so `claude-zen --permission-mode acceptEdits` and friends all work.
**The `/model` picker won't list these models.** Switching means quitting and relaunching with a different `-m`. That's a Claude Code limitation.
Why a proxy is needed
The obvious setup — pointing `ANTHROPIC_BASE_URL` straight at `https://opencode.ai/zen` — does not work.
Zen's Anthropic-compatible `/v1/messages` endpoint mangles tool schemas when it translates them to OpenAI format, dropping `function.name`:
Failed to deserialize the JSON body into the target type:
tools[0].function: missing field `name`
The models themselves do tool calling perfectly — the identical request to `/v1/chat/completions` returns a correct `tool_calls` response. Only the Anthr
Related Skills
Agrout Bridge
Local HTTP proxy that routes OpenAI-/Anthropic-compatible clients (OpenCode, Claude Desktop, Cursor) through A
Openapis
Free Claude API + free OpenAI API in one unified proxy. Drop-in compatible with Anthropic SDK, OpenAI SDK, Cla
Gemini Claude Bridge
new Gemini-to-Claude protocol converter for using Gemini models as Claude Code backend. Fixes 3 LiteLLM bugs
Agentsw
CLI provider switcher for coding agents: manage OpenAI/Anthropic-protocol providers in one place and sync them
Claude Code Provider Switching Kit
Switch Claude Code between Anthropic, GLM, and OpenRouter (including free models) with one command.
Free Claude Code
Nemo Code — FREE Claude Code alternative. NVIDIA open models + Claude Code CLI framework. One command install.
Related Agents
Fleet Architect
Fleet Architect of the fleetsmith fleet for Meta agent-fleet builder: one fleet.yaml spec compiles into coordi
Maker
Reviews OpenCode sessions (shared links or local history), finds repeatable workflows, and proposes automation
Glm Agent
Runs Z.AI GLM on the user's GLM Coding Plan. Three modes - consult (direct API call, 1M context, for inputs to