Agent
An agent that works inside your workspace
The agent reads, writes and acts in your workspace, asks before it writes unless you tell it not to, and cites the sources of its answers.


Sessions, skills and plans
You work with the agent in sessions. Skills describe how your organization does a task, and a plan shows the steps before the agent goes through them.
- Sessions for a task from start to finish
- Skills for work you repeat
- Plans you can read as the agent works
You decide when it writes
For each session you choose between “Ask before writing” and “Write without asking”. With the first, nothing is saved until you approve it.
- “Ask before writing”: you approve each change
- “Write without asking”: the agent saves its work
- The choice is made per session
What the agent can do
The agent can read a page, write a page and search the web. It runs code and analyses in a sandbox, and from the desktop app it can work in a browser.
- Read and write pages
- Search the web
- Run code and analyses in a sandbox
- Work in a browser, from the desktop app
Answers with their sources
Answers cite their sources, so you can open the page or record an answer rests on and check it yourself.
- Citations that open the source
Agents in depth
These pages go deeper into agents that work together, connected systems and procedures.
Give the agent a first task
Start with a page of your own, or book a demo and we show it with your questions.
