How I Got Claude Code Running on My Gemini Subscription (Without Any API Keys)

TL;DR: I built a lightweight local proxy called CodeWeaver that translates Claude Code's Anthropic API calls into Google Cloud Code Assist requests — using your existing Gemini CLI OAuth session. Zero dependencies, one file, ~500 lines.

The Problem

If you're a developer working with AI coding agents, you've probably noticed a split in the ecosystem:

Claude Code has arguably the best UX among AI coding CLIs — great streaming, tool use, multi-turn conversations, and a polished workflow.
Gemini CLI comes bundled with your Google account and doesn't require managing API keys, but the CLI experience is different.

I found myself wanting the Claude Code interface with my Gemini subscription. Paying for two AI subscriptions felt wasteful, and setting up API keys for every tool was a pain.

So I built a bridge.

The Architecture

CodeWeaver is a local HTTP proxy that sits between Claude Code and Google's APIs:

Claude Code ──> Anthropic API request ──> CodeWeaver ──> Gemini Request ──> Google
                                                                                │
Claude Code <── Anthropic response <── CodeWeaver <── Gemini Response <────────┘

It's a single Node.js file using only native modules (http, fs, path, crypto) — zero npm dependencies. It runs on localhost:8099 and speaks the Anthropic Messages API on one side and the Google Cloud Code Assist streaming API on the other.

How Protocol Translation Works

This was the most interesting part. Claude Code and Gemini have different API shapes, and bridging them required handling several translation layers:

1. Message Role Mapping

Anthropic uses "user" / "assistant" roles, while Gemini uses "user" / "model". Simple enough. The tricky part was handling tool interactions.

2. Tool Use ↔ Function Calling

When Claude Code calls a tool, it sends a tool_use content block. CodeWeaver translates this into Gemini's functionCall format:

Anthropic:            Gemini:
tool_use block  ──>  functionCall { name, args }
tool_result     ──>  functionResponse { name, response }

The trickiest bit was preserving the thought signature — Gemini 3 models include a cryptographic signature in tool responses for security. CodeWeaver encodes this inside the tool_use ID with a custom encoding scheme (toolu_{name}_{random}_sig_{sig}) and decodes it on the way back.

3. Schema Sanitization

This was the biggest gotcha. Claude Code sends OpenAPI 3.1-style tool schemas with keywords like $schema, anyOf, oneOf, const, and exclusiveMinimum. Google's API strictly rejects these.

CodeWeaver's sanitizeSchema() function strips unsupported keywords, flattens anyOf/oneOf into their typed variant, and converts const to enum arrays — all while preserving the actual type structure.

4. Streaming Response Translation

Claude Code expects Server-Sent Events (SSE) with a specific event structure:

event: message_start
event: content_block_start
event: content_block_delta    (text_delta / input_json_delta)
event: content_block_stop
event: message_delta
event: message_stop

Google sends a different SSE format. CodeWeaver maintains state about the current active content block (text vs. tool_use) and translates Google's stream chunks into the Anthropic event sequence on the fly.

Authentication: The Best Part

No API keys. No .env files. No vertex AI setup.

CodeWeaver reuses the OAuth credentials from your existing Gemini CLI session. When you run codeweaver --login, it:

Spins up a local callback server on port 8085
Opens a Google OAuth consent screen (PKCE flow)
Exchanges the authorization code for access + refresh tokens
Optionally provisions a free-tier Cloud Code Assist project
Stores credentials locally

After that, it handles token refresh transparently — keeping your session alive without any manual intervention.

Rate Limiting and Fallbacks

Google's Cloud Code Assist has strict per-minute quotas on Pro models. CodeWeaver handles this with a self-healing fallback:

If a Pro model (gemini-2.5-pro) hits a 429, the proxy transparently retries with gemini-2.5-flash
If the Flash model also hits limits, it parses the retry-after duration and waits
Up to 3 retry attempts before giving up

This means your Claude Code session rarely fails — it just gracefully degrades to a faster (but still capable) model when traffic is high.

Model Mapping

Claude Code has a hard-coded model selector. CodeWeaver maps those to their Gemini equivalents:

Claude Code Model	Gemini Model
Sonnet (default)	gemini-2.5-pro
Haiku	gemini-2.5-flash
Opus	gemini-3.1-pro-preview
Fable	gemini-3.1-flash-lite

You can also pass raw Gemini model IDs directly: claude --model gemini-3.1-pro-preview.

The Result: 500 Lines, Zero Dependencies

The entire proxy is a single proxy.js file. Here's the breakdown:

~80 lines — OAuth login flow with PKCE
~60 lines — Schema sanitization
~120 lines — Request translation (Anthropic → Gemini)
~200 lines — HTTP server + streaming response handling
~40 lines — Token management and config

Total: ~500 lines of vanilla Node.js. No TypeScript, no frameworks, no npm packages.

What's Next

CodeWeaver is intentionally focused on the Gemini → Claude bridge right now. But the roadmap aims for something bigger:

Plugin architecture — provider-agnostic backend swapping
Ollama support — run Claude Code against local models
OpenAI / DeepSeek / OpenRouter — any API-compatible provider
Auto failover — cascade through providers on rate limits or outages
Docker image — containerized deployment
Web dashboard — visual provider configuration

The vision is to make Claude Code a universal frontend for any AI model backend.

Try It

curl -fsSL https://raw.githubusercontent.com/whysooraj/CodeWeaver/main/claude-gemini-proxy/install.sh | bash
codeweaver --login
codeweaver
claude

Or clone and run manually:

git clone https://github.com/whysooraj/CodeWeaver.git
cd CodeWeaver/claude-gemini-proxy
node proxy.js --login
node proxy.js
claude

Requirements: Node.js 18+, a Google account, and the Gemini CLI.

CodeWeaver is open source under MIT. Contributions, issues, and feature requests are welcome!

Links:

GitHub: github.com/whysooraj/CodeWeaver
Install: One-liner above or clone the repo

Have questions? Found a bug? Want another provider supported? Let me know in the comments or open an issue on GitHub.