All Blog

A local team of bots, or one cloud assistant?

Big tech gives you one all-purpose assistant. Teebots takes another path: several role-clear bots on your own computer, bring your own key, login optional.

A local team of bots, or one cloud assistant?

Big tech usually gives you one all-purpose assistant. Teebots takes another path: you keep several bots with clear jobs on your own computer, bring your own key, and you can start without a login. This post maps three paths, and when a local team fits better.

What you are missing is often not a smarter model

People switch assistants when the current one writes drafts, looks things up, and edits contracts in the same thread until the context muddies and the output goes soft.

I hit a version of this myself. I used OpenClaw for an AI news monitor. It worked well at first. Then I loaded many skills and MCPs, asked it to write articles and retrieve news in the same place, and it got confused while the token bill kept climbing. The problem was not that the model was too weak. It was that one worker was asked to do too many jobs at once.

If what you want is division of labor, not another chat box that claims to do everything, a local multi-bot setup starts to make sense.

There are roughly three paths

Cloud personal assistants. ChatGPT, Claude, and many big-tech agents sit here: one session does everything, switching devices is easy, and compute lives on their side. The tradeoff is that the workspace and sign-ins stay tied to the cloud, and roles blur into "a little of everything."

Self-hosted or open-source stacks. Hermes, OpenClaw, and similar tools put models, tools, and deployment under your control. They suit people who will operate the stack. You can keep adding skills and MCPs, and then you get the failure mode above: one agent writing and searching at once, with a bill that is hard to contain. The capability is there; you have to design the constraints.

Local desktop multi-bot. That is where Teebots sits. Named bots work together in a group chat on your computer. You @ someone, or they hand work to each other. Requests use your own model keys (BYOK). The app itself is free, and login is optional. Sensitive steps such as submitting a form, paying, or entering a password stop and ask you first.

None of the three is always best. If you need to switch devices anytime, or you want someone else's always-on cloud PC, the cloud is simpler. If you want deep self-hosting and accept ops work, an open-source stack fits. If the files, browser, and local logins already live on this machine, and you want work split by role, a local team is the right tool.

What a "team" looks like in practice

In Teebots you can open a group and add several bots. The one that writes code can use Claude. The one that opens public pages can use another model. The one that writes the public note can use a third. If you name nobody, a bot with a public bio may answer, or nobody may. When it is unsure, it asks you with a choice card.

The site already sketches a one-person company split: CTO edits code, CEO sets direction, COO turns work into a checklist, CMO writes the announcement. Creators can split scout, copy, analytics, and competitor watch. The point is not cute role names. Each bot keeps its own memory and tool scope, and the others cannot read its private memory.

That is the opposite of hanging twenty skills on one assistant. You draw the boundary on people (bots), not on an ever-growing plugin list.

Why local and BYOK matter

Chats, memory, files, and browser logins stay on this computer. Model requests go straight to the provider you pick, not through Teebots servers. The app is free. Model spend shows up on your key bill, so you can see which bot is burning tokens.

Downloads today cover macOS (Apple silicon and Intel), Windows (x64 and ARM64), and Linux (x64 and ARM). Android and iOS are on the way and are not shipping features yet.

When not to pick a local team

If you are already deep in a big-tech plan and need the same cloud PC across devices, a local team will not solve that. Cloud multi-bot products such as Grok Bot fit people who already live in the Cursor / platform ecosystem. We compared "where it runs, how models are chosen, and whether you pay another seat" on Teebots vs Grok Bot. That page is meant to help you match, not to trash the other product.

If you enjoy running Hermes or OpenClaw yourself, that is fine too. Teebots is for a different job: you do not want to stand up a stack first, and you do want a small team on your machine with clear bot roles.

How to start

  1. Open teebots.ai/download and install the build for your system.
  2. Add your own model key.
  3. Create two or three roles with short job descriptions.
  4. Open a group, assign work with @, and watch who burns tokens and who muddies context.

The app is free. Model fees stay on your provider bill. That is the number to watch.

Your local team of bots

Not locked to one model vendor. Each bot uses your own key — Claude, OpenAI, DeepSeek, Kimi, or any compatible endpoint. Requests go straight to the provider.