Claude Code
Build, push and deploy from your terminal-based agent, then let it roll back if a release goes wrong.
AI & MCP
Connect any AI tool to your hosting
Zinn Digital® exposes your hosting as a single MCP server, so your preferred AI agent can manage, build and deploy sites in plain language. Connect Claude Code, Claude Desktop or Cursor once, and every tool we add becomes available to it automatically.
- 700,000+sites hosted worldwide
- 1connection for every AI tool
- 99.99%Uptime assurance
- Per toolscoped, revocable access you grant
One connection, every AI tool
MCP (the Model Context Protocol) is the open standard for connecting AI agents to real systems. We support the protocol, not each app, so you connect the tool you already use and control hosting conversationally.
Cursor
Read logs, purge caches, change DNS and deploy from your editor without leaving your flow.
Any MCP client
Any client that supports remote MCP servers with a header connects the same way. ChatGPT is the exception for now: it only signs in with OAuth.
Claude Desktop
Diagnose a site, prepare a migration or check billing usage in natural language, from the desktop app — connected through the free Zinnector® CLI.
Three primitives your agent can use
The MCP server is a thin adapter over our engine API, so it reuses the same action catalog, RBAC and audit trail. It exposes three things.
Tools
Actions the AI can invoke — the same API endpoints behind the dashboard, with no separate logic to drift out of sync.
Resources
Read-only context — your account, each site's summary and its health — so the agent reasons with real data before it acts.
Prompts
Reusable workflow templates we publish, such as "diagnose this site" or "prepare a migration", so common jobs are one instruction.
A tool catalogue for real hosting work
Every tool is gated by the connected identity's scopes. A sample of what your agent can do:
Because the MCP server rides the same API, everything new we add to the platform becomes available over MCP automatically — you are never waiting on a per-tool integration.
- List and create sites, deploy them, and roll back a release
- Purge a site's cache and choose which crawlers may index it
- Take a backup, and restore a site from one
- Sync a site's files with its git repository, in both directions
- List your domains and edit their DNS records
- Read a site's error and access logs, build logs, metrics and usage
- Check index status and SEO metrics
- Read plan, usage and invoices, and buy add-ons with confirmation
Built for AI acting on live infrastructure
An agent with access to production needs guardrails that are designed in, not bolted on. Ours are:
Token, org and RBAC scoped
Your AI client holds a connection token scoped to your org; every session is bound to your user and permissions, with Postgres row-level security enforced. A connection can never out-scope the person who created it.
Per-tool, revocable tokens
Tokens are scoped and revocable per tool, with sandbox and production kept separate, so you grant only what each agent needs.
Destructive-action confirmation
Delete, suspend, billing and large-spend actions require explicit confirmation or a human-approval policy before they run.
Spend caps and full audit log
Rate limits and spend caps bound AI-triggered paid actions, and every MCP action — identity, tool, arguments, result — is audit-logged.
FAQ
Which AI tools can I connect?
Any MCP client that can send a token. The dashboard gives you the exact config for Claude Code, Claude Desktop and Cursor, and any client that supports remote servers with a header connects the same way. ChatGPT is the exception for now: it connects only by OAuth sign-in, so it cannot use a token.
How do I connect a tool?
In your dashboard, choose "Connect AI tool" to create a connection — the server address and a token — then paste the config the dashboard gives you into your agent. Tokens are scoped, per-tool and revocable, with separate sandbox and production access.
Can an AI agent do something destructive by accident?
No. An agent can only do what the connected identity is permitted to do, and destructive actions — delete, suspend, billing or large spend — require explicit confirmation or a human-approval policy. Spend caps and a full audit log bound the rest.
Does the agent see real data before it acts?
Yes. It can read a site's error and access logs, its metrics and usage, and its health and configuration, so it diagnoses with real context before taking any action.
Is there a cost to try it?
You can start on a card-free 14-day trial, and every plan carries a 30-day money-back guarantee, free migrations and no vendor lock-in.
Related
Give your AI agent the keys, safely
Connect the AI tool you already use to hosting built for scale, with scoped access and a full audit trail. Start on a card-free 14-day trial.
Start freeWordPress plugins your AI agent can work with
Each of these plugins makes your WordPress site an MCP server, so Claude, Cursor, VS Code and other AI assistants can do real work on it, signed in as you and limited to what you are allowed to do.