Conifer

Developers using AI coding tools must choose among many harnesses (Claude Code, Codex, Cursor, Zed, VS Code, etc.), hundreds of models (frontier labs and open-weights like Qwen, DeepSeek, GLM, Kimi, Llama), and various providers, then manage caching and routing themselves. Picking the right combination for each query is manual and fragmented, and switching between cloud, self-hosted models, and local hardware adds further complexity. Today a developer has to send queries to a single fixed setup rather than getting the optimal harness-model-provider-cache combination for each individual request.

Conifer provides a free runtime and router installable via a single curl command on Mac, Windows, or Linux, plus a unified API gateway (api.conifer.build) giving access to over 200 models from frontier labs and open-weights providers, or the option to bring your own provider keys. For every incoming query, Conifer dynamically selects a harness, model, provider, and cache strategy, described as 'routing to the Pareto frontier,' so the developer only sends the text of their request. It works with cloud models, self-hosted models, and local hardware through any llama-compatible endpoint or Conifer's own engine, and cloud tokens bill at the model's own rate with no markup.