
A step-by-step guide for people who have never written code.
You describe the task in plain words — the AI builds the app and deploys it to a secure server on its own.
It's a way to create software by describing the task to an AI in words — instead of writing code by hand. The AI writes the program. You set the direction, review the result and refine it.
The AI writes the code. From you — a clear explanation of the task, like briefing a new teammate.
What needs to happen, for whom, what data, and what counts as the result. The more specific, the better.
The AI writes the code, creates a server and publishes the app. You review and refine.

Two different docs: one to learn it yourself, the other to hand to an AI model.
A detailed walkthrough — follow it point by point, step by step.
Open the platform and click "Sign in with Bitrix24". Authorize (a QR code works too) and pick the right account from the list.
A key is the pass your app uses to work with your Bitrix24. There are two ways to create one:
In the "Create API key" dialog, set:
Click "Create". The key is shown ONCE — copy the "Copy prompt" text right away (it already includes your key) or "Copy key".
Open Claude Code or Codex, paste the copied prompt and add, in plain words, what app you need. See what a prompt looks like below.
The AI writes the code, creates a Black Hole server and deploys the app. If it lacks permissions, it will tell you which to add — go back and edit the key.
Go to the "Black Hole servers" section, find your server and click "Open". If an embedding is configured, the app shows up right inside Bitrix24 — in the menu, a CRM card, a task or a chat.
My API key for the Vibecode platform: vibe_api_xxxxxxxxxxxx Documentation: https://vibecode.bitrix24.com/v1/me Build me an app [describe the app] and deploy it to a server
For yourself and a single app. Works on your behalf — everyone sees the same data. Created in the "API keys" section.
When a team uses the app or it needs to be embedded inside Bitrix24. Each employee gets their own permissions. The "Authorization keys" section.
The key is shown once. You can't view it again — copy it right away. Lost it or closed the window? Just create a new one and revoke the old.
Treat the AI like a new teammate: give context, discuss the idea and review what's done. The clearer you explain, the closer the first result is to perfect.
What to do, for whom, what data to use and what counts as the result. "A dashboard by week", not just "a dashboard".
Before asking for code, talk the idea through with the AI in plain words — like a product manager, not a programmer. And ask the model to ask questions: each question exposes a future bug.
Not "everything at once". In small steps, a "prompt → review → refine" loop. Fix things surgically rather than rewriting the whole project.
"Build a deals-digest app for the sales team. Every morning at 9:00, send the manager a chat summary for yesterday: total of new deals, top-5 reps by revenue, deals without tasks and deals stuck longer than 7 days. Take the data from CRM, the "Sales" pipeline; amounts with the change vs the day before. If there were no new deals — send a short "no deals yesterday". Add a "Refresh now" button. Deploy it to a server and send me the link."
It has it all: what the app is, for whom, when it fires, what data and in what form, the edge case and the finish — deployment. The fuller the paragraph, the closer the result is to perfect on the first try.
Your first app isn't the finish line. It's easy to improve, embed into Bitrix24 and roll back if a change didn't land.
Go back to the AI, describe the edits in plain words — it changes the code and re-publishes the app.
Show the app where it's handy: in the menu, a CRM card, a task or a chat — pick the embedding points.
The platform keeps a code version on every publish (the "App sources" section). If an edit broke something, you can re-publish a previous version.
Once you're comfortable with the first app, the cabinet has a lot more to offer. Every section hands the AI ready-made code via a "Use" button.
Your home screen after signing in: balance, today's requests, success rate and quick entry into every section.
Your app's calls: how many requests, the success rate, where the errors are. Shows up after the first real calls — an empty section right after deploy is normal.
Balance, top-ups and the charge history. Servers and powerful AI models consume credits; at zero they stop.
Chatbots for Bitrix24 messenger — they reply to messages and react to account events.
A ready AI-employee bot working inside Bitrix24 (the "Agents" section). Access is granted on request.
File storage for apps — pay per use, isolated per key. The "Storage" section.
Connecting external services: the platform issues a regular revocable key. Approve only the sign-in you started yourself.
Server backups: a snapshot of the whole disk, restore into a new server. Enabled in settings.
Dense hosting — several apps on one server. For those scaling their solutions.
If you're a Bitrix24 account administrator, you can configure who does what on the platform.
Two separate settings: cabinet access (sign-in) and the right to create keys, servers and bots. Configured independently.
How many keys, servers and bots one employee can create — set by the account administrator.
An action log (who, when, from where) and handling access requests from employees — in the administrator sections.
If you "don't have permission" for some action, the administrator most likely set it up that way. Message them or send an access request right from the cabinet.
A few simple rules protect you and your customers' data. Here are the three main ones — read the full breakdown on the security page.
One step stops 99% of account-theft attempts. Turns on in a minute.
It opens everything you have access to. Don't put it in code, don't forward it in chats.
It's full internet access without a sign-in. Don't enable it unless you understand the consequences.
When the app is ready, ask the AI to review the generated code for vulnerabilities (best with a "fresh look", in a new session). After deploy, repeat the check: production may surface risks that weren't there during development.