Before sending a codebase to an AI model, you have no idea if it fits the context window or how much it'll cost. You either overpay with a premium model for a simple task, or hit context limits mid-session.

ModelRouter analyzes your files locally (no upload, no server), estimates token count, detects complexity, and recommends the optimal model for your specific task — with real cost estimates.

If your workspace is too large for any model, it builds an automatic segmentation plan: splits by module, assigns the right model per segment, and shows total cost.

No signup, no API key, single HTML file. Drag your folder and go.

Would love feedback on the recommendation logic — especially if the model suggestions feel off for your use case.