---
title: The web terminal's operation log — what is logged, why, and who can read it
nav: Terminal log
summary: The web terminal shows Operation log recording at the top. The Agent J host keeps a text log of the lines you send and the text output — no screen recording, no camera, no sound. It stays on the Agent J host, is never uploaded, and is deleted after 30 days; if something goes wrong, you and Agent J can trace it back.
order: 61
nav_order: 50.5
---
# The web terminal's operation log — what is logged, why, and who can read it

This page covers the operation log of Agent J’s full terminal and compatibility line mode on ordinary Agent J hosts. The Olares app uses Olares’ terminal, which does not use the logging or masking rules here; see [Olares web terminal](/docs/web-terminal/#olares).

While you use the [web terminal](/docs/web-terminal/), its top always shows **Operation log recording** followed by the time left in this session. Click it to open this page.

## What is logged {#what}

- The full terminal keeps masked output and terminal size changes, without a separate keystroke record. Line mode also keeps each line you send.
- The text output the terminal shows, the same as you see on the page.
- When each part happened, so you can read it back in order.

Before anything is logged, the Agent J host hides common passwords and keys, the same way as on the page. The Agent J host login password is never in it. In line mode, a line sent to a running program is kept only as a placeholder, not its text. The full terminal does not separately save raw input. Text a program echoes is output and is masked before it is logged.

It is a text log only: no screen recording, no camera, no sound.

Masking reduces accidental disclosure; it cannot recognize secrets in every encoding or unusual format. Do not print passwords, keys or tokens, or share terminal screenshots containing them.

There is also an activity log: which device opened and closed the web terminal and when, plus the terminal lines and sudo calls you confirmed. That one keeps fingerprints only (a hash of each line), never the command text.

## Why it is logged {#why}

The web terminal runs commands right on the Agent J host, so it is powerful. The log is there so a problem can be traced:

- **Trace back.** Something broke or a file went missing: you and Agent J can go through the log line by line and see what was done on the Agent J host at that time.
- **Misuse.** If someone else ever gets hold of your remote, what they did in the terminal is logged too: you can see which device, when, and what.

## Where it is kept {#where}

Only on the Agent J host, in Agent J's data folder (by default `~/.local/state/agentj/terminal/casts/`), one file per session. Only your account on the Agent J host can read the files. The log is never uploaded anywhere.

## Who can read it {#who}

- **You.** At the Agent J host, this shows the recent activity log and the file name of each session's log:

```bash
agentj terminal logs
```

- **Agent J.** It can read the lines and output of the current session, or of the last one for up to 30 minutes after it closed, to help you carry on with a problem; this is masked the same way first. Older logs are in Agent J's data folder, which the Agent cannot read on its own; for it to look at one, you hand that file to it at the Agent J host.
- **Us.** The log is never sent to us automatically and never goes into a support report or diagnostics by itself. When you want our help, you decide whether to send anything and which part: a support message the Agent writes goes out only after you say yes, and it is masked once more on the Agent J host before it leaves.

## How long it is kept {#keep}

Each session's log is kept for 30 days. Older ones are deleted the next time the web terminal opens. The copy Agent J can read directly is kept in memory only, for at most 30 minutes after the close, and is gone when Agent J stops or restarts on the Agent J host.

## Why 15 minutes at most {#time-limit}

Each time the web terminal opens it lasts at most 15 minutes by default, then closes by itself. This is so a forgotten terminal does not stay open: if you walk away from the browser, the terminal does not wait there for anyone. It also closes after 5 minutes with no input or output in the full terminal, or no line in line mode, and when the browser has been on another tab or minimised for more than 60 seconds. A short disconnection is retained for at most 30 seconds. Only the same device can resume after a fresh passkey check; the original expiry stays unchanged.

On the original remote, choose **Verify and resume terminal** and use your passkey to return to the same process. The page shows the remaining time. Unsubmitted input on the old connection can require a new terminal. Choose **Verify and open again** while you still have access. If temporary access expired, choose **Prepare an access request**, review the draft in chat and send it for the owner to approve.

For longer sessions, set `--minutes` when you switch it on at the Agent J host, at most 60; see [Switch it on at the Agent J host](/docs/web-terminal/#enable). When the time is up and you want to go on, open it again; every open needs Face ID.
