Turning a 14-bot office into a web page: Grok Bot Pages pulls outputs out of chat

Akshay demonstrated Grok Bot's Pages feature on X: bots can publish their work as accessible pages instead of leaving it in chat logs. The original post gave three examples: a live HQ office for an organization made up of 14 bots, a custom analytics dashboard, and an interactive learning guide. The video is below.

Of the three examples, the most immediately understandable is the 14-bot HQ. A chat window is good for follow-up questions and trial and error, but not for displaying work outputs. With many bots, outputs get scattered across the conversation stream, making it hard to review, share, and compare them. Pages puts the results under a single URL, which at least solves the question of "where is the thing?" For readers without a development background, you can think of it as a publishing page for bots: the bots do the work, and the page carries the results.

But "live" does not mean real-time collaboration. A commenter asked whether the page updates automatically or requires the bot to refresh it. Akshay replied that he set up a routine that fetches all updates and updates the page every hour. In other words, the page refresh frequency is hourly, not every write being pushed to the front end. Another commenter followed up: when 14 bots update the HQ at the same time, how are state conflicts handled—does the page support concurrent writes, or is it last bot wins? This is a crucial question, but the original post did not elaborate. Packing multi-bot collaboration into a single page makes the presentation layer easy; state consistency is not easy.

Cost was another issue in the replies. Someone asked how much 18 bots cost per month and whether it is cost-effective compared with humans. The original post gave no numbers. From the material alone, it is impossible to judge value for money, because cost depends on model call volume, task frequency, and the complexity of the routine. The page itself does not produce answers; it just lays out the bots' outputs.

Akshay also showed a custom analytics dashboard and an interactive learning guide. The former is good for viewing metrics in one place; the latter is good for turning learning materials into interactive pages. Both directions point to the same thing: a bot's output should not serve only a single conversation; it can accumulate into a reusable page. A commenter said that building a UI for agents helps with management—he himself spends most of his time in the terminal, but he understands that a UI has value for others. This feedback is more practical than "cool." Terminal users and page users are not the same group.

There was also a video reaction in the replies:

Pages can solve presentation, but not conflicts and cost in collaboration. The latter two issues were raised in the replies, and the post did not provide complete answers. If you want to evaluate Grok Bot Pages, first check whether your bot output needs to be reviewed and shared repeatedly, then check whether an hourly update frequency is acceptable. If both conditions hold, the page has value; if you only want real-time collaboration, look for another tool.

Original post: Akshay's X post