Hosting your AI can run.
Belov Cloud is built to be operated by AI — ours and yours. Our copilot keeps every site healthy on its own. Yours — Claude, ChatGPT, Cursor, Claude Code — connects in a minute and works through the same tools you have: deploys, logs, WP-CLI, staging, env vars, backups. You choose how much it may do; anything it can't undo waits for your yes.
Plain words in, real work out.
No dashboards to learn and no scripts to write. Say what you want; the agent picks the tools — and shows its work.
One URL. Approve once.
Belov Cloud speaks MCP, the open standard AI apps use for tools. Paste the server URL, sign in, choose the access, and it's on.
claude.ai & Claude Desktop
Settings → Connectors → Add custom connector, paste https://belov.cloud/api/mcp, approve it on Belov Cloud. Done.
ChatGPT
Add a custom MCP connector (developer mode) with the same URL, then approve it. ChatGPT gets the same tools Claude does.
Claude Code
claude mcp add --transport http belov https://belov.cloud/api/mcp — it signs in on first use. In a terminal it can also use SSH and WP-CLI directly.
Cursor & VS Code
Add https://belov.cloud/api/mcp as an MCP server in Cursor or VS Code — both sign in on first use. Scripts and CI use the REST API or npx belov.
Powerful enough to help. Fenced so it can't hurt.
You approve it once
Each app asks on Belov Cloud before it gets anything. You pick the account, and full access or read-only — the app can't widen that later.
It's you, never more
An agent acts as the person who approved it: their role, their sites. A developer limited to two sites connects an agent that sees exactly those two.
Asks before the irreversible
Pushing staging to live or restoring a database needs an explicit confirmation — and the apps themselves ask you before running anything marked destructive.
Staging first, by habit
Plugin updates and search-replace can run on a staging copy first, and pushing to live backs live up before it. PHP switches and deploys are health-checked: a broken one never takes traffic.
Everything on the record
Every change lands in your activity log with who did it — "ana@studio.com via Claude" — next to what our own copilot did.
Off in one click
Account → AI agents lists every connected app. Disconnect, and its access is gone that second; remove a teammate and theirs goes with them.
28 tools, one connection.
The same operations you have in the panel. What it can't do through the API — run arbitrary code, raw SQL, edit wp-config — stays on SSH, where a person is in charge.
Look
- sites & details
- activity feed
- runtime logs
- deploys + build logs
- env vars
- backups
- WP-CLI (read)
Ship
- deploy
- wait until live
- roll back
- restart
- set env vars
- switch PHP
WordPress
- WP-CLI
- plugin & theme updates
- search-replace
- cache & page purge
- staging create / refresh / push
Recover
- take a backup
- download link
- restore to staging
- restore the database
- SSH details
And when nobody's asking, our own AI is still on watch — checking every site every minute and fixing what breaks. See it heal a site →
The questions people ask first.
Which AI apps work with Belov Cloud?
Any app that speaks MCP: claude.ai and Claude Desktop (custom connectors), ChatGPT (custom MCP connectors in developer mode), Claude Code, Cursor and VS Code. Scripts and CI use the REST API or the belov CLI, and agents that run in a terminal can also use SSH with WP-CLI.
Is it safe to let an AI change my site?
It only gets what you approve — full access or read-only — and it acts as you, within your role and your sites. Pushing staging to live or restoring a database needs an explicit confirmation, deploys are health-gated with automatic rollback, PHP switches roll back on their own, and every action is written to your activity log. You can disconnect an app at any time.
What can't it do?
Through the API it can't run arbitrary code, run raw SQL, edit wp-config, or install plugins from a URL or zip — those stay on SSH, where a person is in charge. It never gets server-level access, never sees other customers, and can't do more than the person who connected it.
Does it cost extra?
No. AI agent access — the MCP server, the REST API and the CLI — is included on every plan, with no per-call fees. SSH comes with every WordPress site and every full-stack app.
Do I need to be a developer?
No. In claude.ai or ChatGPT you paste one URL, approve the connection, and ask in plain words — "update the plugins on staging and tell me if anything breaks". Developers get the REST API, a CLI and SSH on top.
Can I take the access away?
Yes, instantly: Account → AI agents → Disconnect. Removing a teammate also removes every app they connected, and account API tokens stop working if their creator stops being an owner or admin.
Hand your AI the keys — just the ones you choose.
Move your site for free, connect your AI, and let both of them work.