Which agents beafk runs
The real Claude, GPT, Grok and Kimi, plus one more that does not chat — all on your own subscriptions.
beafk installs and runs the command-line tools their makers ship, signed in as
you. We do not wrap them or replace them with an API. Skills and tool servers
you set once reach every agent you run.
The four you chat with
| Agent | Who makes it | Chat | Signs in with |
|---|
| Claude | Anthropic | yes | your Anthropic account, or an API key |
| GPT | OpenAI | yes | your ChatGPT account, or an API key |
| Grok | xAI | yes | your x.ai account, or an API key |
| Kimi | Moonshot AI | yes | your Moonshot account |
The binaries are the ones their makers ship, under the names they ship them
under: Claude Code, the Codex CLI, Grok Build, Kimi Code.
These four are the ones that appear in a chat. The install puts none of
them on: you add them from the panel, one press each, which is where you sign
them in anyway. Add one today and the rest next month without touching the
machine.
One more, which does not chat
Codex Security (OpenAI) reads a repository and reports what is wrong with
it — codex-security scan .. There is nothing to converse with. It sits with
the others because the questions about it are the same: is it installed, is it
signed in, and how do you sign it in from a phone.
Mixing them
You are not choosing one. A machine can have all of them, a chat names
which one it is with, and a crew can put several to work
on one job at the same time — one scouting, one building, one reading the diff.
They are not equally good at the same things, and you are already paying for
more than one.
What each one costs you
Nothing extra here — every agent is on every plan, and every message goes on
the subscription that agent is already signed in with. What beafk bills for is
the machines on your plan, and a person is never a line on it. See
what it costs.
Next
Signing an agent in is the how, from your phone,
without going back to the server.