model string selects the model and its privacy level, and Nito handles routing, metering, and attestation. New to that frame? Read How the API Works.
- Chat completions: messages in, reply out, at
/v1/chat/completions. - Streaming: the reply as Server-Sent Events.
- Tool calling: function tools the model asks to call; your code runs them.
- Server-side tools: web search, web fetch, and the file parser, run by Nito.
- Reasoning: think first, with effort and summary controls.
- Multimodal: image and document input.
- Model Fusion: several models on one prompt, in the Browser App and Plugin.
- Prompt caching: provider cache markers passed through.
- Safe retries: failed calls are not charged.
- Best practices: latency, cost, keys, logging, fallbacks.
POST /v1/messages and /v1/messages/count_tokens (Anthropic shape) and a stateless POST /v1/responses and /v1/responses/compact. See SDKs and the API Reference.
Related
How the API Works
The request lifecycle.
Chat Completions
The main endpoint.
Privacy Overview
The four levels.
API Reference
Every endpoint.