Auditing Claude Code, Cursor and Antigravity compared

By the AgentReceipt team · Updated 13 November 2026

Teams rarely use only one agent. The agents work differently, so a record has to be built differently for each, but you want one kind of receipt out of all of them.

Claude Code

It supports hooks, so a tool can listen to prompts and edits as they happen. Setup: agentreceipt init.

Cursor

AgentReceipt supports Cursor with its own setup: agentreceipt init --agent cursor.

Antigravity

Antigravity has no prompt hook, so AgentReceipt reads the prompt from the conversation log when an edit, a command or the end of a turn is reported. Setup: agentreceipt init --agent antigravity.

One receipt for all three

Whichever agent was used, the receipt has the same parts: the prompt, the files changed and the flags. That is what lets a team compare work across agents. Set up all of them at once with --agent all.

agentreceipt init --agent all --pre-push