Web Browsing
Your agents can browse the live web during chat and missions. Research competitors, pull transcripts, post to social platforms, fill forms, and take screenshots.
Your agents are not limited to what is in their Brain or the documents you have uploaded. When a task needs live information, or a live action on a site, they open a real browser and do it.
What Agents Can Do on the Web
| Task | Example prompt |
|---|---|
| Research | "Look at what our top three competitors posted on LinkedIn this week" |
| Pull transcripts | "Give me the transcript of this Instagram reel" |
| Scrape content | "Read this article and save the key points to my brain" |
| Verify live data | "Check my Shopify dashboard and tell me today's sales" |
| Post and comment | "Post this to my LinkedIn" or "Reply to the top comment on my last reel" |
| Fill out forms | "Sign me up for this waitlist" |
| Take screenshots | "Show me what this page looks like right now" |
| Navigate multi-step flows | "Book a call on their Calendly for Tuesday afternoon" |
You do not need to tell the agent to "use the browser". They pick the right tool for the task.
Public Sites vs Signed-In Sites
| Type | What agents can do | Needs |
|---|---|---|
| Public pages | Read articles, watch YouTube, check company sites, view public LinkedIn profiles | Nothing. Works out of the box |
| Signed-in actions | Post to your accounts, comment on your reels, pull private Instagram transcripts, read gated dashboards | A saved session from the Vibey Mini extension |
Without the extension, agents still browse the open web. With the extension, they can also act as you on the platforms you have approved: Instagram, LinkedIn, TikTok, X, Facebook, YouTube, and Reddit.
What It Looks Like in Chat
While the agent works, you'll see live progress in the conversation:
- "Opening instagram.com..." the agent started a browser
- "Reading the page..." the agent pulled the content
- "Taking a screenshot" an image appears in chat when done
- "Filling the form..." the agent is interacting with inputs
Once finished, the agent replies with the results: text, screenshots, a saved artifact, or a confirmation the action is done.
Safety and Control
The browser your agent uses is not your Chrome window. It runs in a dedicated, isolated environment Vibey creates per task and destroys when the task finishes. Nothing the agent does affects the tabs you have open.
- Signed-in actions require your consent. The agent only uses a saved session if you've enabled it and you haven't disabled that domain.
- Destructive actions usually pause for your OK. If an action could send a message, submit a form, or spend money, the agent typically asks before doing it.
- Full audit in chat. Every navigation, click, and screenshot appears in the conversation. You can review the whole path.
What Agents Can't Do
- Bypass logins. Without a saved session, the agent sees public content only.
- Solve captchas or 2FA. If a site throws a challenge, the agent surfaces it and asks you to complete it.
- Use sessions you've disabled. If a site is Disabled in your Sessions tab, it's off-limits.
- Use sessions for random websites. Only the platforms on the approved list get session sharing.
How to Use It
Just ask. A few examples:
- "Read this article and pull the three strongest arguments:
<link>" - "Look at the last 10 reels on @competitor_handle and tell me which ones got the most reach"
- "Transcribe this Instagram reel and turn it into five LinkedIn posts"
- "Book a 30-minute call on this Calendly for Wednesday afternoon:
<link>" - "Post this draft to my LinkedIn"
- "Take a screenshot of my competitor's pricing page"
For the richest results on social platforms, install Vibey Mini and enable session sharing. Without it, agents are limited to what's publicly visible.