A local proxy and admin UI that routes coding-agent traffic to 25 cloud or local model providers. It lets developers select, validate, and manage models from a single interface while using compatible coding agents.
Project overview
Developers who need to switch between different cloud APIs or local model servers for tools like Claude Code can use this local proxy to centralize provider management and routing.
Project type
AI Coding · Model Runtime
Use cases
Coding & Development
Deployment
Refer to project documentation
License
MIT
Best for
Developers and AI engineers who manage multiple model API keys or local model servers and need to route coding agent traffic through a configurable local proxy.
Key capabilities
Routes coding agent traffic through a local proxy connected to 25 cloud and local providers.
Choose and validate providers from one local web-based Admin UI.
Allows each coding agent's native model picker to select models exposed by the proxy.
Routes Fable, Opus, Sonnet, Haiku, and fallback traffic to different models.
Runs Claude Code sessions through Discord or Telegram with voice-note transcription using NVIDIA NIM or Local Whisper.
Protects the local proxy with optional token authentication.
Optionally runs Claude Code sessions through Discord or Telegram bots.
Limitations and risks
Local models require sufficient context capacity to accommodate the coding agent's system prompt and tool definitions.
Getting started
Run the curl or PowerShell install script to automate setup, start the server command, open the Admin UI URL shown in the startup log, and validate or apply a provider key in the UI.
Alternatives and comparisons
An open-source code generation model that automatically installs required packages and attempts to run code on demand.
An open-source MoE foundation model series designed for agents that supports frameworks like Claude Code and Cline.
Flagship open-source MoE LLMs optimized for long-horizon agentic engineering tasks.