Claude Code Model Bridge — AI skill for Claude Code
Presents the Claude Code CLI as an OpenAI-compatible model endpoint.
How to install Claude Code Model Bridge
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open dbardi/claude-code-model-bridge and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Claude Code Model Bridge does
Presents the Claude Code CLI as an OpenAI-compatible model endpoint.
Alternatives in AI
- Codex — openai-codex Adversarial code review, Codex CLI integration, cross-model analysis 9.6k ★
- EchoBird — One-click install + model switch:Claude Code,Codex CLI (OpenAI), Grok Build (xAI), DeepSeek Harness, Kimi Code 3.1k ★
- CliRelay — Self-hosted AI gateway for coding CLIs — one OpenAI/Claude/Gemini/Codex-compatible endpoint, with a multi-tena 1k ★
README
claude-code-model-bridge
Presents the Claude Code CLI to any OpenAI-compatible client as an ordinary model endpoint.
Requests arrive as normal chat completions. Each one is answered by running `claude -p`, so the work is billed to a Claude subscription rather than to API credits. The bridge refuses to start when an API key is present, so it cannot quietly spend anything else.
Claude Code's own tools stay off. It acts purely as a model: it requests tool calls, and the client that made the request runs them. An agent harness therefore keeps its own tools, permissions, memory and logging, and needs no changes to use this.
Requirements
- Python 3.13 or newer, and uv
- The Claude Code CLI, logged in (
claude auth statusshould report a subscription) - No
ANTHROPIC_API_KEYorANTHROPIC_AUTH_TOKENin the environment
Running it
uv sync
uv run claude-code-model-bridge
The bridge serves `http://127.0.0.1:8765/v1` and binds to loopback only.
Configuration comes from the environment:
| Variable | Default | Meaning |
|---|---|---|
CLAUDE_BRIDGE_HOST |
127.0.0.1 |
Address to bind |
CLAUDE_BRIDGE_PORT |
8765 |
Port to bind |
CLAUDE_BRIDGE_CATALOG |
the shipped models.yaml |
Path to a catalog file |
CLAUDE_BRIDGE_MAX_CONCURRENT |
4 |
How many calls may run at once |
CLAUDE_BRIDGE_PLUGIN_DIR |
unset | Directory of plugins whose skills the model may use |
Installing it as a service
On a machine with systemd, `scripts/install.sh` installs the bridge as a user service that starts at boot and restarts if it stops.
./scripts/install.sh --check # check prerequisites, change nothing
./scripts/install.sh # install and start the service
./scripts/install.sh --uninstall # stop and remove it
The script installs only the service. It never edits the configuration of whatever client will use the bridge, so a failed install cannot take your assistant offline; p
Related Skills
Routsi
OpenAI-compatible endpoint that dynamically routes each request to the right API model or agent (Devin, Codex,
LLM Bridge
Local OpenAI-compatible LLM API gateway for Google Antigravity (AGY), multi-account Gemini, Claude Code and Co
Bring Your Own Model
Run Claude Code on any OpenAI or OpenAI-compatible model — a tiny, dependency-free local bridge that translate
Local Delegate
MCP that delegates mechanical text tasks to a local OpenAI-compatible LLM endpoint to conserve Claude subscrip
Airoute
Local AI router. One OpenAI-compatible endpoint for every provider you already have, and one key for opencode,
Multimodal Vision Skill
One image-recognition engine for any AI agent — 3 protocols (OpenAI Chat/Responses, Anthropic Messages), multi
Related Agents
Codex CLI
Codex CLI → any OpenAI-compatible endpoint
Openrouter Agent
Runs a model from any of ~60 vendors through OpenRouter's OpenAI-compatible endpoint, billed per token against
Fable5 Apfel Engineer
Heavy-lifting Fable-5 engineer for the apfel project (Apple on-device FoundationModels CLI + OpenAI-compatible