Connect 12 coding agents to your preferred models using free tokens
This tool acts as a bridge between coding agents you already have and various AI model providers. It allows you to combine your desired models and tools without being locked into a specific platform.
Flexible combination of agents and models
It supports 12 coding agents, including Claude Code, Codex, and VS Code Chat. You can connect models from 61 providers, such as NVIDIA NIM, OpenRouter, and Groq. You can assign different models to the Fable, Opus, Sonnet, and Haiku tiers of Claude Code to fine-tune cost and performance. All connections are handled locally and designed to comply with provider terms of service.
Extended usage environment from terminal to browser
You can use this tool via terminal commands, desktop apps, IDE extensions, mobile messengers, and browser-native sessions. Discord and Telegram support voice memos, which are converted to text using local Whisper or NVIDIA NIM. The browser offers multi-agent, multi-model sessions to maintain the same coding experience in a web environment. This multi-platform support allows you to continue coding regardless of your work environment.
Built-in features for stability and token efficiency
Several stability features are built in to allow coding to continue without interruption due to provider outages or rate limits. These include automatic retries, configurable rate limits, concurrency control, and automatic fallback for models and providers. Additionally, it uses RTK and Headroom technologies to reduce token usage by up to 90%. These features ensure efficient use of free credits and continuity of work.
Installation and initial setup process
Installation is available via the curl command on macOS and Linux, and via a PowerShell script on Windows. After installation, running fcc-server opens the management UI, where you enter your API key and select a model. If you use NVIDIA NIM as the default provider, you simply register the API key generated from build.nvidia.com.
Limitations and precautions for free credits
This tool is an independent open-source project and is not affiliated with Anthropic. The availability and limits of free tokens vary by model provider and may be restricted at any time due to policy changes. Therefore, you should check the latest terms and usage limits of the provider before important tasks. When using local models, you must select a model with sufficient context length to handle the agent's system prompt and tool definitions. The license is AGPL-3.0-only, and you must comply with its terms for commercial use.
