An audit trail for
everything your agents do.
When an agent does something you have to explain later, terminal scrollback is not an answer. Pushary records every approval, denial, and automatic decision, append-only and exportable, across every agent you run.
Mis en avant sur Hacker NewsA record that outlives the session
A log you cannot search, export, or read six months from now does not really exist. This one does.
Append-only by design
Once an interaction is recorded it cannot be edited or deleted. The only change allowed is writing your answer to it once.
Every decision, including the silent ones
Not just what you approved. The actions your policy auto-approved are logged too, and those are the ones you will want to look up later.
One trail across every agent
Claude Code, Codex, Cursor, and Hermes write to the same record, so you do not stitch together five different logs to see what happened.
Searchable and filterable
Filter by agent, tool, decision, or time. Find the one command you are looking for without scrolling a terminal you already closed.
Exportable
Pull the trail out as CSV or JSON to hand to your team, attach to an incident, or keep for a compliance review.
Who, what, and when
Each entry records the action, the arguments, the decision, and who made it, so the record actually answers the question later.
Start keeping the record.
Connect an agent and every decision lands in the trail from the first run.
Frequently asked questions
What is an AI agent audit trail?
It is a record of every action your AI agents took and every decision made about those actions: which command ran, who approved or denied it, and the ones the policy decided automatically. Pushary keeps that trail across all of your agents.
Is the audit trail tamper-proof?
It is append-only. Once an interaction is recorded it cannot be altered or removed. The only permitted change is recording your answer to a question once, so the history you read is the history that happened.
Does it log the actions that were auto-approved?
Yes, and that is the point. The decisions you never saw, because the policy handled them, are exactly the ones you will need when you are reconstructing what an agent did. Pushary records them alongside the ones it asked you about.
Can I export it?
Yes. The full filterable trail and CSV or JSON export are on the Agent Pro and Team plans. On Free and Agent you get a bounded recent-activity view.
Does it work across different agents?
Yes. Claude Code, Codex, Cursor, Hermes, and any MCP client write to the same trail, so you have one record instead of one log per tool.
How much does it cost?
Paid agent plans start at $9.99 per month with a 7-day trial. The full filterable audit and export are on Agent Pro and Team.