Say what you want. Agent builds it and runs it.
Describe your app in a conversation, in Persian or English. Agent writes the code, creates the database, cache and bucket it needs, and deploys every change as a preview with its own address. It asks you before production or anything that costs money.
What you get
A conversation in Persian or English
Write the way you would talk to a developer. Conversations are kept, and Agent remembers the earlier messages. When something is unclear, it asks a question and waits for your answer.
On this platform's products
Agent deploys the app or function and creates the SQL database, cache, bucket, queue and secrets it needs in the same project. What it builds is an ordinary project you see in the dashboard.
A preview for every change
Every change is a preview with its own address, beside the conversation. Try the result right there and say what to change.
Undo any step
Every step Agent takes is a version. If you do not like the result, roll back to the earlier version with one click.
Your approval before what matters
A production deploy, a custom domain, a paid resource and a wallet top-up wait for your approval. Approve or deny, with a note for Agent if you like.
A key for one project only
Agent works with an API key that reaches only that project. It cannot touch your other projects, the team or billing. Turn Agent off and its key is revoked at once.
Voice messages
Hold the microphone button and speak. The voice message's text lands in the message box, so you read it, fix it if needed, and send it.
Attach any file
Send a screenshot, a mockup, a PDF, a CSV file or an archive with your message. Agent sees an image as it is, so its layout and look reach the app.
See what it does
Thinking, editing files, running a command or searching the web: every step shows live. Tell Agent something while it works, or stop it.
The code is yours
Export the app's code, or carry on with the same project in any other coding agent through the MCP server. Agent never locks you in.
The same checks as any build
An app Agent builds passes the same security checks as any build before it goes public. Agent reads build and runtime errors and fixes them itself.
See what each message costs
Each message's cost in credits shows beside it. Agent's daily usage is in the project's settings.
In the dashboard, or through the API
The same conversation works in the dashboard and through the API: send a message, follow every step live, and approve or deny what waits for you.
# turn Agent on for the project (owner or admin)
curl -X PATCH -H "Authorization: Bearer $KEY" \
-d '{"enabled": true}' https://api.ibel.ai/v1/projects/shop/agent/settings
# the first message starts a conversation and a run
curl -X POST -H "Authorization: Bearer $KEY" \
-d '{"text": "یک فروشگاه ساده با سبد خرید و صفحهٔ مدیریت بساز"}' \
https://api.ibel.ai/v1/projects/shop/agent/conversationsFor developers: the model API and the agent runner
If you write code yourself, what runs under Agent is yours to use. Call language models with this platform's key and address, and run your own agents in an isolated environment, paid in tomans from the same wallet.
- A model API in the two common chat formats, so existing client libraries work by changing the base address
- A key per project, each with its own spending cap
- Run your own agents in an isolated environment, with schedules and workflows
- Any coding agent deploys here through the MCP server with a project key: previews first, production on approval
Paid by the work each message does
Each message to Agent uses credits for the work it does. Credits come from the month's credit first, then from the wallet. The apps Agent builds pay for their own products like any other project. The price of a credit is on the pricing page.
Questions
No. Write what you want, try the preview and say what to change. If you do code, read the code, export it, or carry on with your own tools.
Anything in a preview: writing code, deploying previews and setting variables. A production deploy, a custom domain, a paid resource and a wallet top-up always wait for your approval. Agent cannot git push, make keys, or reach the team or billing.
You do not have to. Agent has one model, which we chose and tested on Persian, so the same app costs the same each time and you only think about what to build.
The apps, databases and files Agent builds run on our servers inside Iran, like any other project. Agent's own work, your messages and the code it writes, is processed outside Iran. If that route is cut, Agent pauses until it is back, and every app it built keeps serving.
Work in progress stops and Agent's key is revoked at once. What it built stays in the project, and the conversations carry on when you turn it back on.
Build your first app with Agent
One project, one message, one preview.
Sign-up is by invite for now. If you have a code, create your account.
Sign up withEmail codeSMS code