Free Gemini 3 Pro API Tutorial: Hands-On Integration Guide for Claude Code & Cherry Studio

Google's Gemini 3 Pro model is impressively capable, but many developers don't realize that with an open-source tool, you can call Gemini 3 Pro entirely for free via API — not only for coding assistance but also for image generation with Imagen. Even better, a single API endpoint supports OpenAI, Gemini, and Anthropic protocols simultaneously, enabling seamless integration with mainstream AI tools like Claude Code, Gemini CLI, and Cherry Studio.

This tutorial walks you through the tool configuration step by step and validates the results through three practical scenarios.

Core Tool: Anti-Graffiti Tools Introduction & Configuration

How It Works

If you've used Google AI Studio, you probably know that Google provides free usage quotas for models like Gemini 3 Pro and Imagen (image generation) on the platform. Google AI Studio is Google's AI model experimentation platform for developers, where users can interact directly with the Gemini model family through a web interface. Google provides each account with daily free call quotas (Rate Limits), typically including per-minute request limits and daily total token caps. This strategy aims to let developers experience the latest model capabilities at zero cost while controlling server costs through quotas.

The core concept behind the open-source project Anti-Graffiti Tools is: converting API requests into page requests after logging into a Google account through the browser. In other words, when you call a model via API, it actually consumes the free quota from the AI Studio web interface, achieving zero-cost access to Gemini 3 Pro.

Installation & Configuration Steps

Step 1: Download and Install

Go to the project's GitHub Release page, download the latest version (e.g., 3.33.2 Windows edition), double-click the installer, and follow the prompts to complete installation.

Step 2: Add Google Account

After opening the software, first add a Google account. The default method uses OAuth authorization (the first option), which is the simplest approach. OAuth (Open Authorization) is an open-standard authorization protocol that allows third-party applications to access user resources without obtaining the user's password. In this scenario, the OAuth flow redirects users to the Google login page for authentication, then Google returns an Access Token to the tool. The tool uses this token to make requests to AI Studio on behalf of the user, which is more secure than storing usernames and passwords directly since tokens can have expiration periods and permission scopes, and users can revoke authorization at any time in their Google account settings.

To get more free quota, you can add multiple Google accounts for rotation, or log in with a paid Google Pro account for unlimited calls.

Step 3: Obtain API Connection Information

After successful login, enter the API panel where you can see the service port, API key, and supported model list. The right side of the panel displays the Base URL and API Key — these two pieces of information are crucial for integrating with various AI tools.

Step 4: Configure Model Mapping (Optional)

Since different AI tools may define model names differently, you can use the model mapping feature to map model names from other software to the corresponding Gemini versions, ensuring calls work correctly. For example, Claude Code requests a model named claude-sonnet-4-20250514 by default. Through the mapping feature, this can be automatically converted to gemini-2.5-pro, with the middleware handling the format conversion transparently to the caller.

Step 5: Start the Service

Once all configuration is complete, click the "Start Service" button in the upper right corner, and the API service is officially running.

Technical Background: Multi-Protocol Compatibility

Understanding Anti-Graffiti Tools' multi-protocol compatibility requires knowledge of the differences between current mainstream AI APIs. OpenAI's Chat Completions API uses a messages array with role/content structure; Anthropic's Messages API requires x-api-key and a specific version number in request headers; Google's Gemini API uses a generateContent endpoint with a parts structure. Multi-protocol compatibility means the middleware needs to parse requests from different formats, uniformly convert them to the format the target model understands, and then convert responses back to the structure the caller expects. This adapter layer design is very common in API gateways and proxy services, enabling a single backend service to serve various tools across the ecosystem.

Hands-On 1: Calling Gemini 3 Pro in Claude Code

This is a particularly clever approach — instead of using Gemini's CLI, you call Gemini 3 Pro directly within Claude Code. This is possible because Anti-Graffiti Tools fully implements the Anthropic API protocol specification.

Claude Code is a command-line coding assistant from Anthropic that allows developers to collaborate with AI in the terminal through natural language for code writing, debugging, and refactoring tasks. It supports SPEC (Specification-Driven Development) workflows, where technical specification documents are generated first before implementing code step by step. Claude Code calls Anthropic's Claude models by default, but since it communicates through standard API endpoints, the underlying model can be replaced with any compatible service by providing an Anthropic-compatible Base URL and API Key.

Configuration Method

Just like using any other API proxy service, you can switch via cs switch or third-party tools like Web, simply filling in the Base URL and API Key. You can also manually set the corresponding environment variables in the settings.json file.

Test Results

After configuration, launch Claude Code and switch the model to Claude Sonnet 4.5 (which actually calls Gemini 3 Pro under the hood). Enter the instruction: "Write a 3D particle animation that supports gesture-based zoom and rotation." You can see it enters the SPEC (specification development) workflow, with the model responding normally and generating code.

Returning to the Anti-Graffiti Tools dashboard and refreshing the quota, you can see the values have changed, confirming that Claude Code successfully called the Gemini 3 Pro model.

Hands-On 2: Native Gemini Model Calls via Gemini CLI

Using Gemini CLI to call Gemini 3 Pro is arguably the most "authentic" way to use it — the experience is virtually identical to the official one.

Gemini CLI is Google's command-line AI development tool, positioned similarly to Claude Code and GitHub Copilot CLI. It allows developers to interact directly with Gemini models in the terminal environment for code generation, project initialization, file editing, and other operations. The advantage of CLI tools is direct access to the file system and project context, enabling AI to understand the entire codebase structure and make global modifications. By configuring API endpoints through environment variables, developers can flexibly switch model backends without modifying their workflow.

Configuration Method

Simply set the environment variables for API Key and Base URL. Important note: If you don't have a global proxy enabled, it's recommended to manually set proxy environment variables to ensure smooth network connectivity.

Test Results

After launching with gemini, the experience is completely consistent with normal usage. For example, entering /init to initialize the current project quickly generates Gemini's MD documentation.

A major highlight of Gemini 3 Pro is its frontend UI development capability. You can provide UI prototypes and have it develop a Flutter APP project based on the mockups — the entire process is very smooth. Gemini plans a complete development roadmap, and you can continuously push it forward until the entire project is complete. This multimodal understanding capability (processing both image and text inputs simultaneously) makes the conversion from design mockups to code unprecedentedly efficient.

Returning to the dashboard page again, you can see the usage changes for the corresponding models, confirming successful API calls from Gemini CLI.

Hands-On 3: Unlocking Chat & Image Generation in Cherry Studio

Cherry Studio is a popular AI chat client that, combined with Anti-Graffiti Tools, can unlock both Gemini 3 Pro's conversational abilities and Imagen's image generation capabilities.

Imagen is a text-to-image generation model series developed by Google DeepMind, based on diffusion model architecture. Similar to Stable Diffusion and DALL-E, Imagen transforms random noise into high-quality images through a gradual denoising process. Its core advantage lies in deep understanding of natural language prompts, benefiting from Google's accumulated expertise in large language models. Imagen 3 is the latest version in the series, with significant improvements in image quality, text rendering, and detail representation, particularly excelling at generating photorealistic images and artistic style works.

Configuration Method

1. Open Cherry Studio, go to system model settings, and find the Gemini configuration entry
2. Add both Gemini 3 Pro (chat model) and Gemini 3 Image (Imagen image generation model)
3. Copy and paste model IDs directly from Anti-Graffiti Tools' model list

Chat Capability Test

Create a new window in the chat box, select the Gemini 3 Pro model, and enter "Create a 3D particle effect sphere using a single page." The response speed is very fast, and a 3D particle sphere effect is generated quickly with support for zooming in to view details.

Although results might not match AI Studio's performance without a system prompt configured, you can continue iterating. For example, asking it to "change particles to circles with an overall sphere shape" immediately improves the result. It's recommended to add custom system prompts to fully leverage Gemini 3 Pro's capabilities. A System Prompt is instructional text preset at the beginning of a conversation that defines the AI's behavior patterns, output format, and domain expertise — essentially assigning the model a "role" that significantly improves output quality for specific tasks.

Image Generation Test

Open a new chat window, switch to the Imagen model, and directly enter an image description prompt. Testing shows it can generate not only comic-style landscape scenes but also sequential series of images one by one, with overall satisfying results.

Important Notes for Free Gemini 3 Pro API Usage

Through the open-source tool Anti-Graffiti Tools, you can call Gemini 3 Pro and related advanced models at zero cost, flexibly integrating with various AI tools including Claude Code, Gemini CLI, and Cherry Studio. Key advantages summarized:

• Completely Free: Consumes Google AI Studio's free page quota with no credit card required
• Multi-Protocol Compatible: Simultaneously supports OpenAI, Gemini, and Anthropic API protocols
• Multi-Scenario Coverage: One-stop solution for coding assistance, AI chat, and image generation
• Highly Extensible: Supports multi-account rotation and model mapping for flexible adaptation to different needs

Note that Google AI Studio's free quota has daily limits. According to Google's official documentation, the free tier for Gemini 2.5 Pro typically limits to 5 requests per minute and 25 requests per day (specific values may change with policy updates). This is generally sufficient for light usage and development testing but may hit limits during intensive development. If your usage volume is high, consider adding multiple Google accounts for rotation or upgrading to a paid Google Pro account for unlimited call quotas. Additionally, network environment is crucial — since Google services require proxy access in some regions, ensure proxy settings are correct to maintain stable API connections.

Finally, it's worth noting that this approach of converting page quotas to API calls carries certain compliance risks. Google's Terms of Service may restrict automated access behaviors, so users should carefully read the relevant terms before use and primarily limit usage to personal learning and development testing purposes.

Key Takeaways

• Anti-Graffiti Tools enables free access to Gemini 3 Pro by converting API requests into page requests
• A single API supports OpenAI, Gemini, and Anthropic protocols simultaneously, integrating with multiple mainstream AI tools
• Supports three major use cases: Claude Code for coding, Gemini CLI for project development, and Cherry Studio for chat and image generation
• Free quotas have daily limits; expand usage through multi-account rotation or upgrading to a paid Pro account
• Model mapping feature resolves model naming differences between tools