Every capability on this page is included on every plan. We don't tier-gate diagnostics, backups, observability, or compliance — the product is the product. The plan tier picks resource size, not feature set.
Evidence
trace 4e9a82c: browser → nginx → php-fpm → mysql
slowest span: SELECT * FROM cart_items WHERE cart_id = ?
19,840 rows · 14.2s · no index used Code Location
app/Http/Controllers/CheckoutController.php:127
$items = $cart->items()->with('product')->get(); Recommended Fix
1. Add index: cart_items (cart_id, created_at) 2. Eager-load product.images to drop N+1 Expected p99: 14s → ~120ms
Detect, diagnose, report, recommend, contain. Five steps, in that order. Other hosts kill first and explain never. We diagnose first, capture evidence, and contain only when there's no other option.
Evidence
trace 8c7e3a1: stripe webhook → handler::process exception: Undefined index "payment_intent" occurred: 18 times in last 24h · 0.4% of webhooks side effect: 7 orders left in "awaiting_capture"
Code Location
app/Webhooks/StripeHandler.php:84 $intent = $payload['payment_intent'];
Recommended Fix
1. Guard with array_key_exists() before access 2. Verify Stripe signature before parsing payload 3. Replay the 18 affected events from the audit trail
Included on every plan.
100% trace capture. No sampling. Per-tenant cost attribution from day one.
ClickHouse-backed unified store: metrics, logs, traces, business events, and cost data all SQL-queryable in one place.
Included on every plan.
A restart destroys the evidence. So we diagnose first — capture the forensics, find the root cause, then take the least destructive action — and we hold both your application and the platform underneath it to that rule. The difference is who pulls the trigger: on your app, you do; on our own infrastructure, we do.
Most hosts kill first and explain never. You get a ticket that says "high CPU detected, we killed your pod" — no diagnosis, no root cause, no record of what went wrong. We do the opposite: nothing is restarted until the diagnostic state is captured and written down.
Your application: you hold the switch.
When something in your app degrades, we diagnose it and tell you exactly what we'd do to fix it. Whether we actually do it is your call — with one exception.
The platform underneath: it heals itself to the same standard.
The infrastructure your app runs on — the proxies in the request path, the databases, the deploy engine — follows the same diagnose-first discipline. Because it's our infrastructure, we act on it without waiting for you. Least destructive first:
Our own databases never get a generic drain-and-restart: we kill specific queries, fail over to a verified-ready replica, and make one change at a time. The destructive actions — a failover, a restart — stay behind a human approval gate. Permanently.
Included on every plan.
Same APIs, same data, same access as human operators. Point your coding agent at our MCP server and it drives the platform from your IDE.
Two MCP servers. External for your developer agent (Claude Code, Cursor, and the rest), internal for our ops team. ~35 tools at launch.
Included on every plan.
Push to GitHub. Or SSH in and rsync. We accept both.
Code is built into an immutable image, deployed blue/green, and gated by health + canary metrics before traffic shifts. Failed deploys auto-revert with full forensics preserved.
synsmarts deploy for workspace-backed tenants.True zero-downtime deploys — not a best-effort rolling restart. The new build comes up beside the live one and takes zero production traffic until it clears its health and canary checks. The cutover is atomic: requests in flight finish on the old slot while new requests land on the new one. Because each slot runs its own object cache, the new slot is already warm — no cold start, no thundering-herd cache stampede the moment you go live. Sessions ride a shared session store across the switch, so a customer mid-checkout never gets logged out or loses their cart. And if the new build fails its gates, traffic never moved in the first place — the deploy auto-reverts and your customers never saw it. No maintenance window, no "be right back" page, no dropped requests.
A real staging environment, not a toy. Spin up a second instance at the smallest tier and you get a true staging environment: the full production stack, real AWS, real diagnostics — not a stripped sandbox that behaves differently from prod. Deploy the same git SHA to both instances to promote a verified build, and run the diagnostic report against staging before it ever reaches your customers.
Multi-model code review on your deploys. Prism is the review gate we run on our own platform changes: every change goes in front of a panel of independent frontier AI models that have to reach consensus before it ships. One reviewer has blind spots — a panel that must agree catches what any single model misses. Opt your deploys in and the same gate reviews your code before it reaches the build.
Migration safety: we run it before you do. Most hosts read your migration files and guess whether they're safe. But schema migrations and data patches execute arbitrary code — file content alone can't tell you what a Magento Data Patch will do to your data. So we don't guess. When a deploy touches the schema, we clone your live database, run the migration against the throwaway clone, and classify what actually happened before a single byte of production data is touched.
Included on every plan.
Fast is the default, not an add-on. Every tenant serves from the edge, with the PHP and cache layers tuned before you deploy a line.
The same diagnostic engine that names a slow query also keeps the fast path fast: static and cacheable responses are served from Cloudflare's global edge, dynamic responses come off a pre-tuned PHP stack, and the object cache survives every deploy so a release never cold-starts your site.
Included on every plan.
Backups are table stakes, not a premium tier.
Hourly XtraBackup. Continuous binlog shipping at 1-minute intervals. Hourly EBS snapshots. Tamper-proof cross-region copies. Same retention for every tenant.
Included on every plan.
PCI-DSS SAQ D as a Service Provider. GDPR co-equal.
CDE boundary published, audited annually, validated quarterly via ASV scans and continuous file-integrity monitoring at the container layer.
Included on every plan.