This project provides a reverse proxy server for IDE APIs, specifically supporting Claude Code, Codex CLI, and Gemini CLI. It addresses the need for managing multiple API accounts, automating token refreshes, and handling API quotas, offering enhanced reliability and functionality for developers integrating AI models. The primary benefit is a centralized, robust interface for interacting with various AI providers, simplifying complex API management tasks.

How It Works

KiroProxy acts as a unified gateway, translating requests between different AI API protocols (OpenAI, Anthropic, Gemini) and the target clients. Its core design emphasizes multi-account management, allowing for load balancing and session stickiness across multiple user accounts. It features automatic token refreshing before expiry, health checks for account availability, and request rate limiting to mitigate account suspension risks. Advanced capabilities include full support for tool/function calling across all supported protocols, image understanding, and web search integrations.

Quick Start & Requirements

Users can opt for pre-compiled binaries from the Releases page or build from source. To run from source: clone the repository, create and activate a Python virtual environment, install dependencies via pip install -r requirements.txt, and then execute python run.py. The proxy can be accessed at http://localhost:8080 (or a specified port). A Web UI is available for monitoring, logging, and configuration, alongside a comprehensive Command Line Interface (CLI) for account management, login, and server control.

Highlighted Details

• Multi-Protocol & Tooling: Full compatibility with OpenAI, Anthropic, and Gemini APIs, including robust support for tool/function definition, invocation, and result handling.
• Advanced Account Management: Features multi-account rotation, automatic token refresh (15 minutes prior to expiry), account health checks, and automatic disabling of suspended accounts.
• Enhanced AI Capabilities: Integrates image understanding and web search tools for Claude Code and Codex CLI clients.
• Reliability & Context Management: Implements request rate limiting, automatic error retries with exponential backoff, and sophisticated dialogue length management (summarization, truncation, caching) to maintain context and prevent errors.

Maintenance & Community

No specific details regarding maintainers, community channels (like Discord/Slack), sponsorships, or a public roadmap were found in the provided README content.

Licensing & Compatibility

The project explicitly states it is for "learning and research only" and prohibits commercial use ("本项目仅供学习研究，禁止商用"). This restriction significantly impacts its suitability for enterprise or commercial applications.

Limitations & Caveats

The most significant limitation is the strict prohibition of commercial use. While the proxy includes features to manage dialogue length limits (e.g., summarization, truncation), the underlying API input length constraints still apply. The project appears to be under active development, with frequent version updates.