One governed endpoint
Every request is authenticated, checked against a budget, sent to a model you chose, and recorded.
Panorama
An AI gateway that runs in your own infrastructure and connects coding-AI spend to the work it paid for. One endpoint, one container, no prompt or code leaving your network.
Pick where to start
Section titled “Pick where to start”I want to understand itWhat happens to a request, in one diagram, and the three things Panorama will never do.
I want to deploy it on AWSOne CloudFormation template, single sign-on, and Claude Code pointed at the gateway.
I want to deploy it somewhere elseDocker Compose and SQLite on a server you run.
My company runs it and I write codeGet a token, then configure Claude Code, opencode or Codex CLI.
What it does
Section titled “What it does”Cost you can trust
Input, output, cache read and cache write are priced separately, per model and per region. A coding agent is mostly cache reads: price them as input and the bill is overstated several-fold.
Cost per branch, ticket and PR
Read out of the request as it passes, then confirmed against GitHub (experimental). “PAY-1183 cost $18.40, merged.”
Budgets that hold
Per person, per team, per deployment. A soft limit warns and closes the routes you marked as too expensive to keep serving. A hard limit stops the request and says when the budget resets.
Three things it will never do
Section titled “Three things it will never do”- Store a prompt or a response. Bodies are pattern-matched in flight and discarded.
- Show one engineer another engineer’s numbers.
- Hide what it could not attribute. Unattributed spend appears beside every attributed number.
This manual is licensed underCC BY 4.0. Code samples in it are part of Panorama, licensed under theGNU AGPL v3.0 only.