Can you elaborate more on what Grok Bot has (that you are emulating) that isn't available in other agents, and why you are emulating Grok Bot in particular, compared to another agent? What exactly is what "real startup teams need"?
Good question. I think the thing most worth learning from Grok Bot is how the bot interacts with people.
Compare a Claude Code connected to a gateway with Grok Bot. One is a coding agent, the other is a personal agent. When you give Claude Code an instruction, it goes and does it itself: it thinks for a while, works for a while, then gives you a big pile of text. If you say a few more things while it works, your words go into a queue. Whether they cut in or wait, in the end it's the same single agent doing the work, because it's a dev tool, not an assistant.
Grok Bot, as a personal agent, is different. The agent you talk to rarely does the work itself. It's like your assistant: most of the time it only needs to do two things well, talk with you and delegate for you. So when you talk to Grok Bot you don't wait long, you usually get an answer quickly, because the main agent is a lightweight agent tuned only for the conversation. It hands the task to a full-power subagent, keeps an eye on the work and reports back to you. So even when you have several things going and keep sending messages one after another, your messages don't cut in or wait like in a single-threaded queue. Your assistant decides for you whether this message, this task, goes to a new subagent or gets added to the one that's already on it. And you rarely see a Claude Code-sized reply in Grok Bot. Your secretary sorts things out for you. If it's a lot of information, or it needs a special kind of interaction, your secretary uses a page (a special web output), or in-app pop-ups, choice boxes, input boxes and so on. Not only text, and never a wall of text that's hard to read.
I was a long-time user and developer of OpenClaw-like apps before, so I saw right away that this is the biggest difference between a personal agent like Grok Bot and a coding agent connected to a chat app. These are the points I found most worth learning from Grok Bot, and they are the core idea I kept pushing and optimizing for the whole time I built Pots: a personal agent that works like a real coworker / secretary.
All of this is already in Pots, especially the special interactions. Pots didn't build its own app first (that's also why I say Pots cares about what real startup teams need, more on that below). It lets the agent treat each chat app as its own tool and get good at the controls each chat app already has, so inside each chat app Pots gets the same kind of special interactions as the pop-ups and choices built into the Grok Bot app. And since Pots is a local, self-hosted service, I also researched and designed how it can give people a page they can open from outside, the way Grok does: the agent uses Cloudflare and Tailscale services, so its pages can be reached from outside freely and safely.
Now the "what real startup teams need" part. Grok Bot is a separate product, a separate app. If a team wants to use it, everyone can only use it as an outside product and throw their own tasks and their own context at Grok Bot, and the company can't manage it in one place either. Pots is built on top of each team's chat tools. You don't need a separate app, you just bring Pots into your own channels: Slack, Discord, Telegram, Teams and so on. Pots gets all the context you want it to know directly, with no outside tool and nobody passing things along. If Grok Bot is a personal assistant that each person can only use on their own, from outside, then Pots is an in-house team assistant. It's simply a coworker on your team, so internal work takes fewer steps. And because it's built on the team's chat tools in the first place, you get 100% of the features right in your own channels, without a separate app.
Can you elaborate more on what Grok Bot has (that you are emulating) that isn't available in other agents, and why you are emulating Grok Bot in particular, compared to another agent? What exactly is what "real startup teams need"?
Good question. I think the thing most worth learning from Grok Bot is how the bot interacts with people.
Compare a Claude Code connected to a gateway with Grok Bot. One is a coding agent, the other is a personal agent. When you give Claude Code an instruction, it goes and does it itself: it thinks for a while, works for a while, then gives you a big pile of text. If you say a few more things while it works, your words go into a queue. Whether they cut in or wait, in the end it's the same single agent doing the work, because it's a dev tool, not an assistant.
Grok Bot, as a personal agent, is different. The agent you talk to rarely does the work itself. It's like your assistant: most of the time it only needs to do two things well, talk with you and delegate for you. So when you talk to Grok Bot you don't wait long, you usually get an answer quickly, because the main agent is a lightweight agent tuned only for the conversation. It hands the task to a full-power subagent, keeps an eye on the work and reports back to you. So even when you have several things going and keep sending messages one after another, your messages don't cut in or wait like in a single-threaded queue. Your assistant decides for you whether this message, this task, goes to a new subagent or gets added to the one that's already on it. And you rarely see a Claude Code-sized reply in Grok Bot. Your secretary sorts things out for you. If it's a lot of information, or it needs a special kind of interaction, your secretary uses a page (a special web output), or in-app pop-ups, choice boxes, input boxes and so on. Not only text, and never a wall of text that's hard to read.
I was a long-time user and developer of OpenClaw-like apps before, so I saw right away that this is the biggest difference between a personal agent like Grok Bot and a coding agent connected to a chat app. These are the points I found most worth learning from Grok Bot, and they are the core idea I kept pushing and optimizing for the whole time I built Pots: a personal agent that works like a real coworker / secretary.
All of this is already in Pots, especially the special interactions. Pots didn't build its own app first (that's also why I say Pots cares about what real startup teams need, more on that below). It lets the agent treat each chat app as its own tool and get good at the controls each chat app already has, so inside each chat app Pots gets the same kind of special interactions as the pop-ups and choices built into the Grok Bot app. And since Pots is a local, self-hosted service, I also researched and designed how it can give people a page they can open from outside, the way Grok does: the agent uses Cloudflare and Tailscale services, so its pages can be reached from outside freely and safely.
Now the "what real startup teams need" part. Grok Bot is a separate product, a separate app. If a team wants to use it, everyone can only use it as an outside product and throw their own tasks and their own context at Grok Bot, and the company can't manage it in one place either. Pots is built on top of each team's chat tools. You don't need a separate app, you just bring Pots into your own channels: Slack, Discord, Telegram, Teams and so on. Pots gets all the context you want it to know directly, with no outside tool and nobody passing things along. If Grok Bot is a personal assistant that each person can only use on their own, from outside, then Pots is an in-house team assistant. It's simply a coworker on your team, so internal work takes fewer steps. And because it's built on the team's chat tools in the first place, you get 100% of the features right in your own channels, without a separate app.