What's in an agent's reply
Each reply shows which agent and model wrote it, how long it worked, and any changes, links, or next steps it included.
For a task that takes a while, an agent usually sends a short note when it starts and a final reply when the work is done. For a quick question, it may answer right away. Here’s what a reply can contain.
The line above the reply
- The agent’s picture and name, such as Claude Code or Codex.
- The time it was sent.
- How long the agent worked, such as Worked for 3 minutes or Worked for less than a minute, on final replies.
- The model and effort the agent used. Older messages may say Model info not recorded.
- In a shared project, for and a name shows who the agent was working for.
The reply itself
- Formatted text. Agents write in Markdown, a simple way to mark headings, lists, bold text, and code. Enjoy shows it formatted. Wide tables scroll sideways.
- Chips for docs, files, and workflows the agent mentions. Hover over a chip to peek at what’s inside, and select it to open it.
- Links and images. See Links, files, and images in replies for where links open.
- View changes, when the agent saved its work as commits (saved snapshots of your project) during the request. Select it to see what changed in each file. When there’s more than one commit, the button shows how many. See Review the changes an agent made.
After the final reply
- Questions, when the agent needs a decision before it can continue. See Answer your agent’s questions.
- Potential next actions, up to three follow-ups you can send right away. See Use suggested next steps.
- Mark as completed, above the empty message box, once the agent has finished. See Mark a conversation as completed.
To copy a reply, hover over it and select the copy button in its corner. See Copy a message.