See every step the agent takes
Every action the agent takes shows in the reply, with what it read and what it sent, folded into one line once it finishes.
What it includes
Watch it work
A live row names the step the agent is on and counts the seconds once a step runs long.
Failures you cannot miss
A finished run folds into one summary row that opens call by call, and a failed call or a stopped turn shows on the summary itself.
Results you can read
Built-in connectors show results such as a Gmail thread as a readable view rather than raw data.
Charts and apps in the reply
A tool that ships its own view, such as a chart, draws it right in the conversation.
Every step of a script
The tool calls a script made list under its row, so a long job shows what it touched.
The agent asks before it acts
A question, a confirmation, a form, a credential or a connection arrives as a card in the conversation and waits for your answer.
A check before your keys meet web content
After reading a web page, an email or other outside content, the agent asks before using one of your API keys and names where the request goes.
Approve a send before it goes out
Sending an email or a chat message through the agent opens a card naming every recipient and the message first, and runs the identical send once approved.