Documentation

Kalibr

Execution path routing for AI agents. Kalibr automatically picks the cheapest model that works for each task, and reroutes before failures reach your users.

Choose your path

How you integrate Kalibr depends on who is making the routing decisions. Two paths. Pick the one that matches how your system actually works.

i
Not sure which path? If you are writing code or directing a coding agent to write code, that is Path 1. If the agent running your production pipeline is the one deciding what to do next, that is Path 2.

What is Kalibr

Most agents default to one model for everything. That model is usually expensive and chosen before anyone knows which tasks it will actually handle. Kalibr changes this: it observes what requests actually look like, tracks which model succeeds at each type of task, and routes future requests to the cheapest path that is currently working.

When a model provider degrades silently, success rate drops. Kalibr detects this from live traffic and shifts routing before the failure rate spikes. No alert. No rollback. No human required.

This works for text LLMs, voice (TTS and STT), image generation, embeddings, classification, and any model on HuggingFace.

What Kalibr is not: observability (Langfuse, Arize), a model gateway router (LiteLLM, OpenRouter), or a prompt optimizer. It never reads or modifies prompt content. Model calls go directly to the provider.

All documentation