Weave Router cuts coding costs 40%

Weave Router tackles the growing financial burden of frontier models in agentic workflows by replacing manual model toggling with real-time routing. By targeting high-frequency, multi-turn coding agents, it makes long-running development loops economically viable for teams.

• –Uses a local ONNX embedding model to route prompts in under 50ms, avoiding external network latency for routing decisions.
• –Integrates directly as a proxy with support for Anthropic Messages, OpenAI Chat Completions, and Gemini native APIs.
• –Emits OpenTelemetry traces out of the box, allowing developers to visualize decisions and latency in their preferred observability stack.
• –Released as a source-available project under the Elastic License 2.0, providing a self-hosted alternative to closed routing APIs.
• –Addresses the cost surge from tokenizer changes and large context windows in next-generation coding agents.