OrcaRouter, the platform run by Continuum AI, has launched Orca Connect: a feature that turns AI models running locally into APIs reachable from anywhere. Anyone with a personal GPU, workstation or on-premise server can connect it to the platform in seconds, without opening ports on the router or changing firewall rules.
How the connection works
Orca Connect detects the models on the machine automatically and links them to OrcaRouter through a secure connection that starts from inside the network. No public IP address or port forwarding rules are needed. The system supports the most widely used runtimes, including Ollama, vLLM, LM Studio, llama.cpp, SGLang, MLX and TensorRT-LLM.
The exposed models answer through an OpenAI-compatible interface, governed by a single OrcaRouter API key. The same access can be plugged directly into programming tools such as Claude Code, Codex, OpenCode and Gemini CLI.
Sharing and cost control
Sharing happens through revocable links with expiration dates and usage limits, without exposing credentials. The system also offers local-only API keys, policies that decide when the cloud should step in, and spending limits. When the local GPU is busy or unavailable, requests can move to the roughly 200 hosted models on OrcaRouter, but only if the user has explicitly enabled it.
Inference run on the user's own hardware carries no platform fee. Requests still pass through OrcaRouter's servers, while the computation stays on the chosen machine unless cloud fallback is authorised.
A place in the local AI landscape
The move follows a trend that has involved major players in recent months. Nvidia has presented PAIR, an open source tool for linking PCs and Macs into a home cluster, and Google has added local model support to Antigravity. OrcaRouter takes a different route: it does not build a private cluster, but makes the user's own hardware reachable through a single interface that also includes cloud models.
The company presents it as a step toward a more open and decentralised AI infrastructure, in which developers keep ownership of their compute resources. Orca Connect is available now to all OrcaRouter users, with no extra charge for the feature itself.
No comments yet. Be the first!