Chat and agents, ready when you are
AI agents, with nothing to install.
Hosted models are included, your own keys are welcome, and it asks you before anything risky.
Running your own AI workspace is a project
The good open-source workspaces ask for a weekend of setup: containers, ports, model servers that cannot find each other, and a graphics card many people do not have.
And then you hand an agent your files and keys, and hope. When a model only pretends to use a tool, or a web page slips in an instruction, you find out afterwards.
How it works
1. Come in
Sign up with your email or Google. Your workspace and the hosted models are ready when you confirm.
2. Ask
Chat with a hosted model, or add any provider with your own API key. Your keys are encrypted and free to use on every plan.
3. Let it work
Switch to Agent. It searches, reads pages and runs code in a sandbox, shows every step, and stops to ask before anything risky.
What is different
Nothing to install
It runs in your browser, on any computer. Hosted models come with every plan, so you do not need a GPU or a server.
It asks before risky steps
Once a chat has read anything from outside, steps like running code wait for you, showing exactly what will run. Text from the web cannot start an action.
Real tool calls you can see
Agents use the model's own tool calling, and every step appears in the reply. A model that cannot use tools says so instead of pretending.
Your keys, your memory
Bring any OpenAI-compatible provider. Memory you can read, edit and switch off. Private chats stay out of your history and are cleared within the hour.
No surprise bills
Each plan includes a set amount of hosted usage and stops there; it never charges more. Cancel in one click.
Easy to read, everywhere
Clear contrast in light and dark, a layout that works on a phone, and every action reachable from the keyboard.
Pricing
One flat price per workspace, no per-seat charges. Pay yearly and get two months free.
-
Free
For trying it out, or for anyone with their own API keys: those are free and unlimited, plus a little hosted usage each month.
$0
- $1.00 of hosted model usage a month
- Your own API keys, free and unlimited
- Agents, memory, skills and tool servers
-
Personal
For one person working with AI every day: a bigger hosted allowance, agent tools and remote tool servers.
$29 a month
or $290 a year, two months free
- $20.00 of hosted model usage a month
- Your own API keys, free and unlimited
- Agents, memory, skills and tool servers
-
Team
For households and small teams: one shared allowance, one flat price, per-person caps and model limits.
$149 a month
or $1,490 a year, two months free
- $100.00 of hosted model usage a month
- Your own API keys, free and unlimited
- Agents, memory, skills and tool servers
Questions
Do I need a GPU or a server?
No. Everything runs on our side; you only need a browser.
Can I use my own API keys?
Yes, from any provider with an OpenAI-compatible API, on every plan, at no charge from us. They are encrypted, and pages only ever show the last four characters.
Can I use a model running on my own computer?
Only through a public HTTPS address, such as a secure tunnel. A hosted app cannot reach your computer's localhost, and it refuses private network addresses on purpose.
What happens when the included usage runs out?
Hosted models pause until the 1st of the month. Models on your own keys keep working, and we never charge more than the plan price.
Is it safe to let an agent run code?
Code runs in an isolated sandbox, never on our servers. Once a chat has read anything from outside, it waits for your approval and shows the exact code first.
Do you train models on my chats?
We do not. Hosted requests go to model providers under their terms, and requests on your own keys go to your provider under yours.
Can I import my ChatGPT history?
Not yet. It is on the list.
How do I cancel?
One click on the Plan and usage page. You keep the plan until the end of the period you paid for.