Super Amplify Governor

See what your team’s AI coding tools are doing.

Connect a coding agent with one prompt and your report starts filling in within minutes: sessions, tool calls, models, tokens, and what that usage would cost at API list prices. It also tells you plainly what it cannot see.

Free covers one computer with one coding agent. The team evaluation runs 14 days for up to 10 people. Neither needs a card.

The first report

What you see on day one.

This is the report as an administrator sees it, filled with sample data. Yours shows your own numbers, and says “Not reported” wherever a tool does not send a figure.

Your AI footprintSample data

What is connected, what Governor has observed in the last 30 days, and what it cannot see yet.

Connected

9 computers

10 coding agents · 9 people

Activity observed

465 sessions

5,100 tool calls · 94 failed · 16 denied · 1.8 s on average

Model usage

16,430 model calls

973M input · 14.5M output tokens · 28 active days

Estimated API cost

$1,015Estimate

Connected tools at list prices, plus gateway calls priced when they ran. Not what you were billed.
Connected tools, API list-price equivalent:
$612
Gateway calls, priced when they ran:
$402

What is connected

ToolStatusSessionsTool callsModel callsTokensEstimate
Codex4 computersReporting2141,93012,480853.6M$612
Claude Code3 computersReporting1632,41071 failed · 12 denied3,9503,950 through the gateway133.9M$402
Cursor2 computersReporting8876023 failed · 4 deniedNot reportedNot reportedNot reported
Gemini1 computerReportingNot reportedNot reportedNot reportedNot reportedNot reported

Models observed

ModelSeen throughCallsInput tokensOutput tokensEstimate
gpt-5-codexopenaiConnected tool12,480842M11.6M$612
claude-sonnet-4-5anthropicGovernor gateway3,950131M2.9M$402

Not visible yet

  • Whether Codex tool calls succeeded

    Codex hooks report that a tool call finished, not how it ended.

  • Models, tokens and cost for Cursor

    Cursor does not report them, and nothing in Governor adds them yet.

  • Sessions and tool calls for Gemini

    Governor has no activity hooks for Gemini, so it shows as connected and nothing more.

  • Models, tokens and cost for Gemini

    Gemini does not report them. Send Gemini through the Governor gateway to see them.

  • AI used outside connected tools

    Governor sees the agents connected to it and calls through its gateway. A chat app in a browser, or a tool nobody connected, is not in this report.

  • What you were billed

    Dollar figures here are estimates at published prices. Add a provider invoice to compare them with what you paid.

Not collected by this setup: prompts and responses, code and file contents, tool inputs and results, file, folder and repository names.

Each tool call is sent as a fingerprint of the tool’s name, not the name. Each project is sent as a random-looking id made on the computer from a private salt and the folder path.

Two ways in

Start alone, or bring your team.

Both start the same way and take a few minutes. Pick the one that matches what you want to find out.

Connect my own tools

Free. One computer, one coding agent. No card.

  • Paste one prompt into your coding agent and approve the computer in your browser.
  • Your own report starts filling in within minutes.
  • A weekly summary arrives by email. Each one has a link to stop it.
Start free on one computer

Evaluate with my team

14 days, up to 10 people. No card. Needs a work email address.

  • Invite your developers. Each connects their own computer, and you need no coding agent yourself.
  • One report across everyone who joins. It observes and reports; budgets and policy enforcement come with a paid plan.
  • At day 14 you get the findings, and a totals-only summary you can forward.
  • Then Governor returns to Free unless you choose a plan. Nothing is charged.
Evaluate with my team

Rolling out to a whole company, with your identity provider and your own policies? See enterprise AI governance.

What is collected

Activity, not content.

The setup sends counts and timings of what your coding agent does. It does not send what you typed, what the model answered, or what is in your files.

Not collected by this setup

  • Prompts and responses
  • Code and file contents
  • Tool inputs and results
  • File, folder and repository names

Each tool call is sent as a fingerprint of the tool’s name, not the name.

Each project is sent as a random-looking id made on the computer from a private salt and the folder path.

What the setup changes on a computer

  • It installs the Super Amplify connector in ~/.superamplify for your user only. No administrator rights.
  • It adds activity hooks for Governor and a context server for Curator to your coding agent, for every project on that computer.
  • You can remove the computer, or one agent, at any time from Governor or Curator.

What each agent reports

Coding agents do not all report the same things, so the report differs by agent. Other agents can connect for company context and show as connected. See every tool

Codex

Shows

  • Sessions and tool calls from the moment the computer is connected.
  • Models used and token counts, including up to 30 days of earlier usage read from Codex session records on this computer.
  • An estimate of API cost at published list prices, labelled as an estimate.

Does not show

  • Whether a tool call succeeded or failed: Codex hooks do not report it.
  • What Codex was billed: Codex reports tokens, not cost, and a subscription is not billed per token.
  • Usage from other computers, or from session records that are no longer on this one.
Governing Codex

Claude Code

Shows

  • Sessions and tool calls from the moment the computer is connected.
  • How many tool calls succeeded, failed or were denied, and how long they ran.

Does not show

  • Models, token counts and cost: Claude Code hooks do not carry them.
  • Anything from before the computer was connected.

Send Claude Code through the Governor gateway to see its models, token counts and estimated cost.

Governing Claude Code

Cursor

Shows

  • Sessions and finished tool calls from the moment the computer is connected.
  • How many tool calls succeeded, failed or were denied, and how long they ran.

Does not show

  • Models, token counts and cost: Cursor hooks do not carry them.
  • Windows with more than one folder open: the connector skips them rather than guess the project.
  • Anything from before the computer was connected.
Governing Cursor

How it works

Three steps, a few minutes.

  1. Step 1

    Create your workspace

    Sign up with your work account or with email, and name your company.

  2. Step 2

    Paste one prompt

    Pick your coding agent and paste the prompt it gives you. There is no password or key in the prompt; you approve the computer in your browser.

  3. Step 3

    Open your report

    What is connected, what was observed, and what Governor cannot see. Share a totals-only summary when you want someone else to see it.

Governor

See your own report.

Free on one computer, or 14 days with up to 10 people. No card for either, and you can remove a computer at any time.
See what your team's AI coding tools are doing | Super Amplify Governor