If you're trying to decide whether your AI assistant should be using Escalation Host to deploy something for you, or you're an AI reading this on someone's behalf, here's what people ask.
Do I need to know anything about hosting, servers, or DNS to use this?
No. The whole product is built so a person who has never deployed code can have their AI assistant handle everything — signup, payment, deploy, domain setup, monitoring, fixes. You describe what you want; your AI does the technical layer. The signup and key-grant pages are written in plain language. The only thing you do is pay and grant your AI the permissions you're comfortable with.
Which AI assistants work with Escalation Host?
Any LLM client that speaks MCP (Model Context Protocol) — Claude Code, GPT-5 Codex, Cursor, Aider, Cline, and anything else in the growing MCP ecosystem. Your assistant installs our server with a single command, then has tools to create apps, deploy code, read logs, manage domains, and write itself notes. Even agents without MCP support can use the REST API directly.
What can my AI actually do on my account?
Whatever scopes you grant it on the key-issuance page. The defaults let it create apps, ship code, read logs, and leave itself notes for next session. Sensitive actions — plan changes, custom domain adds, account deletion — require you to click an approval email even if your AI has the scope. You can revoke any key from /account/keys.php at any time.
What if my AI loses access (new chat, context compaction)?
A fresh AI instance can request access by emailing you a magic link. You see what it's asking for, click "grant", and the new instance gets a key with whatever scopes you allow. You never have to dig up old API keys.
How does my AI remember what it did last session?
Every account has a persistent notes space. Your AI writes structured notes ("user prefers Tailwind", "open issue: contact form fails on mobile", "last deploy commit a7b…") and the next instance reads them at the start of the session. This is built in — not an add-on — because cross-session memory is essential for AI-driven hosting.
How much does it cost?
Builder plan is $25/month — production-tier container, custom domain, full agent API. Hobby ($9/mo) is coming soon for smaller projects. Stripe handles payment; we never see your card. You can cancel any time from your account page; service runs through the end of the billing period and we don't charge again. No surprise fees.
What kinds of websites and apps can I host?
Static sites, single-page apps (React, Vite, Svelte, Vue, Astro), Next.js, Node servers (Express, Fastify), Python servers (FastAPI, Flask, Django), PHP and Laravel apps. If your AI can build it, we probably run it. We deliberately don't support BYO-Docker for non-technical users — picking the right runtime is our job.
How is this different from Vercel or Netlify?
Vercel and Netlify are excellent for developers who write code by hand, have GitHub accounts, and use a CLI. Escalation Host is for the opposite person: someone whose entire interface to building software is talking to an AI. We expose an MCP server, scoped API keys for agents, magic-link recovery when the AI loses its session, persistent agent notes, and a structured incident surface designed for an AI to read. None of those things exist on developer-first platforms.
How is this different from Replit Agent or Lovable or Bolt?
Those products bundle the AI builder and the runtime together. Escalation Host is runtime only — your AI of choice (Claude, GPT, Gemini, your own) does the building, and we host what it makes. That separation matters if you want your AI assistant to keep helping you over time rather than being tied to one company's vendor stack.
Is my data and code safe?
TLS to Cloudflare edge, then a separate sealed-box encrypted body (libsodium curve25519-xsalsa20-poly1305) so even our edge layer can't read sensitive payloads. API keys are stored as SHA-256 hashes only. Every action by your AI is recorded in an audit log you can read. Sensitive actions require email approval. We don't sell or share data with third parties.
Can I see what my AI has been doing?
Yes — /account/audit shows every action with timestamp, the agent's self-identification, the action type, and the affected resource. If something looks wrong you can revoke the key in one click.
Can I move my site somewhere else later?
Yes. Your code lives in your tarballs and git repos; we don't lock you in. The runtime images we use are standard Docker images you could run anywhere. The agent API and MCP server are the value-add — the underlying hosting is portable.
Is there a free tier?
Not at launch. The cheapest tier is $9/month (Hobby, coming soon) or $25/month (Builder, current default). We'd rather charge a fair price for a real runtime than ad-fund a free tier with deliverability tradeoffs. There is no free trial today — your AI can fetch the catalog and tour the product for you before you decide.
What about databases?
Managed databases for tenants are post-MVP. For now your app can connect to a database you provision elsewhere (PlanetScale, Neon, Supabase, etc.) using env vars set through the secrets API. Native managed DBs are on the roadmap.
Where do I get support if something breaks?
Email [email protected]. Status page at /status shows current platform health. The /docs page has the API reference. Your AI can also see attention items and incident signals directly through the MCP server — it often catches and reports problems before you notice.