How it works

One request. Four decisions. A complete request timeline.

Connect your existing tools to Nautili. One control layer then decides where each request goes, who it belongs to, which policies apply and what gets recorded.

Start for Free Read the docs

Connect

Change the base URL and the key. Keep the SDKs and coding tools you already use. Optional request headers tell Nautili who a request belongs to.

Docs quickstart

Route

Routing profiles per team or key decide which model a request reaches. Fallbacks take over when a model is unhealthy, and Auto Mode only chooses inside the allowlist.

Models

Attribute

Know who generated every request. People are attributed by seat, developer and team; machines by request headers and key owner. Everything lands in one ledger, so every cost has an owner.

Govern

One policy bundle sets budgets, rate ceilings and model and provider access. It is checked in the request path, before anything is sent, and content rules can run in shadow mode.

Security & Deployment

Evaluate

Trial a candidate model beside the incumbent on live traffic, and compare the results before you switch.

Cache

Repeated requests can be answered from cache, on an exact or a semantic match.

Observe

Every request leaves a record: who sent it, which model was chosen, what it cost, which rules applied and how it ended.

req_7f3a9c21e4b82026-10-08 14:32:07

Success

Connectreceivedyour-nautili-gateway
Routeopenai/gpt-4oprimary
Attributeresearchdeveloper seat
Governallowedbudget · rate · access
Cachemisssent to the model
Record$0.0124842 ms · recorded

Illustrative record. The same request ID follows a request from the moment it enters to its final record.

Put every request under control

Connect your stack, define the rules and let Nautili handle the path from request to record.

Start for Free Read the docs