Skip to main content
Maitai is a platform for building, testing, and observing LLM-powered applications. Route your LLM requests through Maitai (recommended), or keep your own client and send Maitai the request/response. Either way, Maitai organizes every call by Application and Intent, runs Sentinels to catch quality problems, and gives you the tooling to fine-tune models and ship changes with confidence.

​
Get started in minutes

Quickstart

Install the CLI, log in, and prompt your coding agent to wire Maitai in.

Base URL integration

Already on an OpenAI-compatible SDK? Point it at Maitai in one line.

Core Concepts

The mental model: Company > Application > Intent, plus Sessions and Sentinels.

API Reference

Prefer raw REST? Every resource, documented.
Everything in Maitai maps to one of three goals. Start wherever your problem is, you can always move to the next pillar as your needs grow.

Build

Improve the model and orchestrate work with fine-tunes, agents, and workflows.

Test

Catch quality problems and prevent regressions with test sets, runs, and criteria.

Observe

See what your traffic is doing across applications, intents, sessions, and single requests.
Start wherever your problem is; you don’t have to go in order. Many teams observe production traffic first, then test it for quality, then build fine-tunes and agents on top. Others come to Maitai specifically to build. For complete, end-to-end recipes, see the Examples.