Show HN: Weave Router Cuts Coding-Agent Costs 40–70% by Routing Each Prompt to the Cheapest Capable Model

workweave/router / Hacker News·medium signal

workweave/router, a model router for Claude Code, Codex, and Cursor, hit 130 points and 81 comments on HN. It embeds each request with a small in-process ONNX model, scores it against frozen intent-cluster centroids, and routes to the lowest-cost model that has historically matched the frontier on that cluster — adding only low-single-digit-millisecond overhead. The team claims 60–70% of Claude Code requests are short, structurally simple completions an open-source model handles at parity, at roughly one-fortieth the cost, installed as a drop-in endpoint change.