If you have not set Browsentic up yet, start with Install, Pair and First run.
Talking to it
| Conversations | Typing and dictating, one conversation per tab, the sessions strip, history |
| Hands-free | Talking to the page through a mic you can drag anywhere, with the side panel put away |
| Instant commands | Commands such as "go back" and "scroll to the top" that run in milliseconds without an agent |
| A-Eye | Pointing at an element instead of describing it, and the agent asking you to point |
| Action cues | A ring on each element the agent is about to click, type into or read |
| Profile | Your details and standing rules, saved once, so forms get real values instead of guesses |
| Blocked sites | Sites no agent or MCP client can read or act on, enforced in the browser where the agent cannot change it |
| Skills | How an instruction is routed to a playbook, and how to write your own |
Acting on a page
| Page actions | The 52 things it can do to a page, grouped by task |
| Screenshots | Viewport, full page or one element, and when a capture is written to disk |
| Theming | Dark mode on a site that has none, and a WCAG contrast audit |
| Captchas | What it will and will not do at a "verify you are human" block |
| Diagnostics | Recording console errors and failed requests to find why a page broke |
| Files | Uploading a file to a page, and capturing a file a page downloads |
Doing things over time
| Monitoring | Watching an upload, a build or a deploy in the background |
| Scheduling | Tasks that run on their own (every weekday at nine, or once tomorrow) and the timers an agent sets for itself |
| Site maps | Teaching it a site once, so later sessions already know the layout |
| Recordings | Doing a repetitive job once yourself, then asking for it "like last time" |
Two pages apply to every feature:
Each tool's exact parameters are in reference/tools.md.