Web Browser
The Web Browser ability lets your AI open public web pages during a workflow conversation. It can read a page, find a field or button, click, type, and press keys. Use it when a task needs information or an action on a website, for example checking a status page or filling in a public form.
Settings
| Setting | Description |
|---|---|
| (none) | Just enable it. No configuration needed |
Before you start
- Your workspace must have the Browser feature turned on. If it is off, the ability does not appear in the Add Ability list, and a workflow that already has it gets no browser tools.
- The ability works in workflow conversations (Assignment Chats).
- The browser acts for the Member in the conversation. That Member's role must allow writing to Chats (
chats:write). A Member whose role has no permissions, such as a patient texting in, gets no browser tools.
Adding Web Browser
Web Browser can be added at either the workflow or task level:
- Open your workflow or task
- Go to the Abilities tab
- Click Add Ability and select Web Browser
- Save
Task-level abilities override workflow-level abilities of the same type.
What to expect
- The browser opens only public web addresses.
- Cookies and site data are kept for the conversation, so a site stays signed in between turns.
- The AI treats page text as untrusted content and does not follow instructions it finds on a page.
- No one can take over the browser in a workflow conversation yet. If a step needs a person, such as a sign-in, a CAPTCHA, or a confirmation, the AI stops and says that step has to be done another way.
- In a test conversation, the AI can open and read pages, but it cannot click, type, or press keys.
Conditional Activation
You can restrict when Web Browser is available using CEL expressions in the ability's Activation condition field. See the Developer CEL reference for variables available in this context, and Abilities for more on conditional abilities.