Skip to content

AI wrote the code. Now there’s a receipt.

AgentReceipt records what Claude Code, Cursor and Antigravity do in a session, flags risky changes with fixed rules, and puts one receipt on the pull request. You review it from one dashboard.

Get early access Works with Claude Code, Cursor and Antigravity
agentreceipt.online
The AgentReceipt dashboard overview, with sample data: sessions recorded, flagged and blocking

The problem

Stop guessing what your agent changed.

An agent can edit dozens of files in one run. Drag the handle to see the difference between reading it all and reading one receipt.

Risk receipt

Claude Code · 3 files · sample

  • Secret-like valueBLOCK
  • Tests weakenedWARN
  • New dependencyWARN
  • Checks switched offWARN

VerdictBlocking

With AgentReceipt

You read one receipt.

+ const res = await fetch(`/api/users/${id}`);

+ const headers = { authorization: "Bearer " + key };

- expect(await fetchUser('2')).toBeDefined();

+ for (let i = 0; i < 3; i++) {

+ try { return await res.json(); } catch {}

+ "p-retry": "^6.2.0",

export async function fetchUser(id: string) {

- return res.json();

+ // eslint-disable-next-line no-explicit-any

+ const cache = new Map<string, unknown>();

+ if (!res.ok) throw new Error(res.statusText);

- test('fetches', async () => {

+ await sleep(2 ** i * 100);

}

+ export const retries = 3;

- expect(res.status).toBe(200);

+ const res = await fetch(`/api/users/${id}`);

+ const headers = { authorization: "Bearer " + key };

- expect(await fetchUser('2')).toBeDefined();

+ for (let i = 0; i < 3; i++) {

+ try { return await res.json(); } catch {}

+ "p-retry": "^6.2.0",

export async function fetchUser(id: string) {

- return res.json();

+ // eslint-disable-next-line no-explicit-any

+ const cache = new Map<string, unknown>();

+ if (!res.ok) throw new Error(res.statusText);

- test('fetches', async () => {

+ await sleep(2 ** i * 100);

}

+ export const retries = 3;

- expect(res.status).toBe(200);

Without a receipt

You read everything, by hand.

  • Every session is recorded

    Prompt, edits and commands, saved in your repository.

  • Fixed rules check the added lines

    No model, so the same diff gives the same result.

  • One receipt per pull request

    A verdict and the evidence, where the reviewer already looks.

How it works

What AgentReceipt does

Four steps, and nothing changes in how you code.

01

Install

Set it up with one command

Install the command-line tool once, then run init in each repository. It adds the hooks your agent needs and a check that runs before you push.

  • Hooks: added for Claude Code, Cursor or Antigravity.
  • Pre-push check: runs before code leaves your computer.
  • Nothing uploaded: sync stays off until you turn it on.
Get early access

02

Record

Work the way you already do

Keep using your coding agent. Each prompt, file edit and shell command is saved to a folder in your repository, and secrets are redacted before they are written.

  • Prompts: what the agent was asked.
  • File edits: what changed, line by line.
  • Commands: what it ran along the way.
Try the rules on a diff

03

Check

Seven fixed rules read the added lines

The rules look only at what the agent added. No model is involved, so the same diff always gives the same result, and each flag points at the evidence.

  • Secrets: secret-like values, redacted in the receipt.
  • Tests: assertions removed, and checks switched off.
  • Scope: new dependencies, sensitive paths and large changes.
Try it on your own diff

04

Review

A receipt on every pull request

When a session ends, it becomes a receipt. With the GitHub App installed, the receipt is posted on the pull request, so the reviewer sees what the agent did and what looked risky.

  • On the pull request: a verdict and the flags behind it.
  • In the dashboard: every session across the team.
  • In git: the receipt travels with the repository.
Request early access

Live demo

Paste a diff. Watch it get flagged.

This is the real rule engine, not a mock-up. Edit the diff on the left and the receipt on the right changes with it.

  • Same engine as the command-line tool
  • Seven fixed rules
  • Nothing you paste is stored
diff · editable
live receiptscanning…

Scanning…

Same rule engine as the CLI · the diff is scanned in memory and is not stored or logged

One dashboard for every session

See which sessions were flagged and what changed, without reading every diff. Managers see the whole team, and employees see only their own receipts.

Get early access
agentreceipt.online/orgs/acme/receipts
The receipts list in the AgentReceipt dashboard, with sample data

Sample data.

Who it is for

Built for the people who ship the code

Pricing

One price. Pay for your team.

No tiers to compare. Everything is included for everyone, and the command-line tool is free.

First 2 days free, no card

$5

per person, per month

Your team

8 people

After the 2-day trial

$40 a month

1100
Start the 2-day trial

Prices in US dollars. Taxes may apply.

Everything included

  • Claude Code, Cursor and Antigravity
  • Every risk rule, and your own policies
  • A receipt on every pull request
  • Dashboard, metrics and receipt export
  • Teammates added by you, each with their own sign-in
  • Receipts kept 7 days, then emailed to the manager and deleted

FAQ

Frequently asked questions

Can’t find your answer? Write to us and a person will reply.

Email the founders

founders@agentreceipt.online

  • $5 per person per month, in US dollars. You only pay for the people in your organization, and every feature is included. Every approved team starts with a 2-day free trial, with no card needed.

  • The person we approve is the manager. A manager sees every employee’s receipts, grouped by employee, adds employees, and can download receipts. A manager does not need an API key. Each employee creates their own key, sees only their own receipts, and cannot download them or change any setting.

  • Seven days. After that the manager is emailed a report with one PDF per employee, and the receipts are deleted from the dashboard.

  • When your request is approved, you get 2 days free with everything included and no card needed. Then the dashboard pauses until you choose a plan. Your receipts and settings are kept, and the command-line tool keeps working on your computers.

  • Yes. Recording what your agent changed, checking it against the rules and building receipts all happen on your own computer and cost nothing. The paid plans cover the hosted dashboard and team features.

  • No. Flags come from fixed rules that read the git diff, and every metric is plain arithmetic over your data. The same input always gives the same output.

  • Nothing, until you run agentreceipt login. After that, each finished session is uploaded with its prompt and changes, with secrets removed first. Set sync to summary-only to send only counts, file names and flags, or to off to send nothing. The server re-checks every upload for secrets and rejects any that look like one.

  • Claude Code, Cursor and Antigravity, each through its own hook adapter. Pick one with agentreceipt init --agent cursor (or all). Gemini models are covered when you use them inside Antigravity, Google’s agent IDE; the standalone Gemini CLI is a different tool and is not wired up. Antigravity has no prompt hook, so its prompts are read from its own conversation log when an edit or command is recorded. Windsurf and other editors are not wired up yet; the adapter interface is agent-neutral, so they can be added without changing the rules or the report.

  • Receipts are written on the author’s machine, so they are self-reported. A missing receipt proves nothing about how code was written. Treat the comment as a review aid and a record kept beside the code, not as tamper-proof evidence.

  • No. It is pattern-based, so it can miss unusual secrets and can flag harmless strings. It is a safety net, not a guarantee. Keep real secrets out of prompts anyway.

  • No. It is scanned in memory by the same rules the CLI uses and the result is returned. The server does not store it, and its logs record only counts, never the diff or the prompt.

See what your agents are changing.

AgentReceipt is invite-only while it is new. Tell us who you are and we will email your sign-in details when your request is approved. Your 2-day free trial starts then.

Questions first? Write to founders@agentreceipt.online

2-day free trial, no card needed. No spam: we only email you about your access. Privacy