You already explained this. Twice. Write it down once instead.
Hook up the AI you already pay for, tell it once who you are and how you like things, and it stays that way — in a chat, on your website, behind an app you built, or in a bot that works overnight. All the AI everyone keeps showing you, without the weekend it usually costs.
Every AI tool starts you from nothing. The same background, the same rules, the same preferences, retyped into every new chat — and because the setup drifts, so do the answers. Configure My AI is where you write that setup down once: pick the AI, pick the model, say how it should behave and what it must never do, drop in the files it should know. After that, every chat, every app, every helper reads the same thing.
If you are new
You are not asked to become a prompt engineer. A Personal Assistant is already waiting. Name it, talk, tap a job when you want one.
If you already live in AI
The vendor is plumbing. Swap Claude for Grok mid-thread. Hit a cap, keep the history. Skills are readable text, so they compose on Ollama the same way they compose on Opus.
Works with ChatGPT, Claude, Gemini, Grok, Mistral, Cohere, DeepSeek and any custom endpoint.
Set it up once.
Never set it up again.
Four steps. About a minute.
Set it up once, use it everywhere
One setup holds the AI you picked, the model, how it should behave, what it must never do and the files it reads from. Talk to it in a browser tab, drop it on your website, call it from something you built, or hand it to a helper that runs on a schedule. It is the same setup every time — so nobody has to remember to paste the same paragraph in first.
Any AI goes in. Everything you make comes out.
A chat window forgets you the moment you close it. A prompt you keep pasting around goes stale the day you write it. This is neither. It sits between the AI you pay for and everything you point it at — and it is the only part you ever have to write.
Change the AI, keep the assistant. ChatGPT, Claude, Gemini or Grok — or any of the others — read the same setup, so switching changes the model and the price and nothing else about how it treats you. That one setup is what answers you in a chat, what talks to visitors on your website, what works overnight on a schedule, and what writes code and puts it live. Change it once and all four catch up.
-
Fix it in one place. Change a rule or swap the model, and the chat, the thing on your website, the overnight helper and the coding session all catch up on their next run. There is no second copy to hunt down, because there was never more than one.
-
It stays up to date. Drop in your policies, your notes, your spreadsheets. Change the file and the answers change with it. Nothing is frozen on the day you wrote it.
-
It respects your no. Say what it must never do, in plain words. That line holds everywhere it turns up, not just where you drew it.
-
It's yours, and it leaves with you. Pay your AI directly with nothing added on top, on keys we keep encrypted. Your chats, your files and your setups are yours to download from day one, whenever you want them.
One place, every major model
ChatGPT, Claude, Gemini, Grok, DeepSeek, Mistral and Cohere all sit behind the same setup. Already have an API account with an AI company? Connect it and pay them their normal rate. Or skip that and use ours. Changing your mind about the model is a dropdown, not a rewrite.
Set it up once. Then point it at anything.
The four on the left are the ones you have seen somewhere — watch one actually run.
A coach who remembers last week
Reads your own logs. Nobody else sees them.
A tutor for your kids
Explains the step. Never hands over the answer.
Someone to answer your website
Your policies, your tone, on a box you drop in.
Something to handle the inbox
Drafts the replies. Email replies wait for your OK.
A brain for the app you built
Your setup behind a web address you own.
Code that writes itself
Say what you want. Review what it wrote.
Your own endpoint, and a real API
Any setup can become a web address of its own, with its own token, rate limit and expiry date — so whatever you built can just call it. The whole thing is written up in an OpenAPI 3.1 specification, so another program can wire itself up without a person reading a word of it. See the developer resources.
What it actually takes off your plate.
Already pay for Claude or Cursor? Spend that, not a second bill.
cma-agent is a small helper for your Mac or Linux
machine. Set it up once and your work runs through a tool you already installed and
signed into — so it comes out of the plan you already pay for, not a new one.
- Your logins never leave your machine. We hold nothing but an address to knock on, so there is no password of yours on our side to lose.
- Your machine calls us, never the other way round. Nothing is left open and there is no firewall to fight.
- It works only in the folders you hand it — real files, real history, and its own copy so nothing of yours is disturbed.
Drives Claude Code Cursor Gemini CLI Antigravity Codex CLI Grok Build CodeWhale Ollama
Install it with brew install 9thdesigns/tap/cma-agent,
then run cma-agent pair and cma-agent start.
The full instructions are on the local CLI page.
Run it on a subscription you already pay for
cma-agent is a small helper you install on a
machine you own. It runs work through a tool already installed and signed in there —
Claude Code, Cursor, Gemini CLI, Antigravity, Codex CLI, Grok Build, CodeWhale or Ollama —
so the work comes out of a plan you already pay for. Your logins never leave your
machine. How it works.
One price here. Your AI bill stays yours.
3 free days and credits to spend, so you can try every AI before you pay anyone. Already have a key? Use it and we add nothing on top. Don't want one? Run on ours at cost plus 10%.
Basic
For one person, or a small thing you run
$18 /month
- 3 active workspaces
- 5 assistants per workspace
- Unlimited endpoints and website chat boxes
- 10 GB storage
- Coding sessions and overnight helpers, included
- Email support
Pro
For a team, or a thing that is growing
$80 /month
- Unlimited active workspaces
- Unlimited assistants and storage
- Shared workspaces for the team
- Skills you can switch on
- Priority support
What it costs to run on ours Pay as you go. Nothing to set up, nothing hidden. Show prices Hide prices
| Provider | Price / 1M tokens |
|---|---|
|
Configure My AI
Default
|
Included |
|
OpenAI (ChatGPT)
|
$4.40 / 1M in · $22.00 / 1M out |
|
Anthropic (Claude)
|
$11.00 / 1M in · $55.00 / 1M out |
|
Google AI
|
$2.20 / 1M in · $13.20 / 1M out |
|
xAI (Grok)
|
$1.38 / 1M in · $2.75 / 1M out |
|
Mistral
|
$1.65 / 1M in · $8.25 / 1M out |
|
Cohere
|
$2.75 / 1M in · $11.00 / 1M out |
|
DeepSeek
|
$1.45 / 1M in · $4.36 / 1M out |
These are the real numbers — our 10% is already in them — and they move when the AI companies move theirs. Bring your own key and you pay them directly, with nothing added.
A minute now. Years of not repeating yourself.
3 free days, credits to try every AI, and everything you put in stays yours to take out.
Getting started
Pick an AI on the home page and you will be talking to your own assistant in about a minute. Every new account gets 3 free days and credits to spend, so you can try all of them before you pay anyone. No sales call, nobody to wait on. See pricing, read the documentation, or see the developer resources.