GEMCLAW
Open Source · MIT License

Run Claude Code
with any LLM

Intelligent proxy that lets Claude Code work with Gemini, Gemma, Llama, or any local model.
9 middleware modules. One command.

View on GitHub
Scroll

How it works

Zero config.
One command away.

01

Install

One npm command. GemClaw registers the CLI globally and sets up the proxy server.

npm install -g gemclaw
gemclaw setup
02

Add your backend

Connect a local model (Ollama, LM Studio, llama.cpp) or add a Google account for Gemini via the dashboard.

gemclaw ui
# → Add Account or configure local model
03

Run

Launch Claude Code through the proxy. It auto-starts, routes requests, and handles everything in between.

gemclaw -g -a
# Gemini + auto mode, fully agentic

9 Modules

Every response enhanced.
Automatically.

CodeLens

Indexes your codebase with BM25 + PageRank, injects relevant context into every request automatically.

Injector

Detects project type — React, Node, Python, Go — and injects framework-specific coding guidelines.

Router

Routes to the right backend, converts Anthropic ↔ Gemini formats, handles multi-account rotation.

Cleaner

Removes model artifacts, garbage tokens, and unclosed code blocks before they reach Claude Code.

Verifier

Syntax-checks every code block in a response before it lands. Catches broken code early.

Guardian

Blocks dangerous shell commands — rm -rf, DROP TABLE, and other destructive operations.

Pinning

Restricts file writes to whitelisted paths via a .gemclaw.json file in your project root.

Loopguard

Detects and breaks infinite tool call loops — stops the model from getting stuck.

Compressor

Strips verbose preambles and unnecessary whitespace to keep context lean.

Backends

Your model, your rules.
Local or cloud.

Local Models

Free · Private · Offline

Any OpenAI-compatible server works out of the box. Run Gemma, Llama, Qwen, Mistral — whatever fits on your machine.

llama.cpp
Ollama
LM Studio
vLLM
Any OpenAI-compatible API
# Start your model server
llama-server -m model.gguf --port 1234

# Run Claude Code
gemclaw

Google Gemini

$22/mo · Gemini 2.5 Pro · Multi-account

Use Gemini 2.5 Pro via Google One AI Pro subscription. OAuth built in — no API keys, no CLI setup. Add multiple accounts for automatic rotation.

Built-in OAuth (no Gemini CLI needed)
Auto-rotate across accounts
Handles rate limits automatically
Gemini 2.5 Pro quality
# Open dashboard → Add Account
gemclaw ui

# Launch with Gemini
gemclaw -g

Get started

One command.
That's it.

Node.js 20+ required. Claude Code installed separately.

Terminal
View source·MIT License·Node.js ≥ 20
G
GemClaw
GitHubMIT License© 2025