Liulum Care monitors any live site every 60 seconds, diagnoses issues in plain English, and fixes managed sites autonomously — often before you notice anything is wrong.
The full repair loop — automated, verified, and documented — in five steps.
Our receptors ping your live deployed URL every 60 seconds — checking uptime, SSL validity, HTTP response codes, and page content markers.
If the heartbeat stops or an error spike is detected, AI runs a fast diagnosis — is it a deployment regression, cache corruption, a DNS issue, or something else?
On managed sites, the repair engine picks the safest allow-listed playbook: rollback, redeploy, clear CDN cache, or restart. Most issues auto-recovered in minutes — the rest escalated fast.
After every fix, Care runs smoke verification. If checks fail, it rolls back automatically — we never leave a broken change live.
Every step — what broke, what we tried, what time it happened, whether it succeeded — is compiled into a timestamped audit and sent via email and WhatsApp.
Most AI tools start cold on every session. Care learns your entire codebase and remembers the context behind every fix, so repairs fit how your project actually works. Your memory is private, AES-256 encrypted, and you can wipe it with one click.
What Care remembers — per repair session
Importance scoring keeps the most relevant context in working memory. Older entries are summarized — never deleted unless you explicitly wipe.
We monitor any site and diagnose in plain language. Auto-fix only runs on what we host or where you've granted access. Honesty is the product.
We watch your site 24/7 and tell you the second it breaks.
Works on: any site
US/UK/UAE pricing per market
We keep your site up — and fix infra issues automatically.
Works on: managed sites
US/UK/UAE pricing per market
Keep it alive AND evolve it — request changes any time.
Works on: managed sites
US/UK/UAE pricing per market
By default, all AI calls are platform-billed on Anthropic models — your data stays inside an Anthropic-only boundary. High-volume clients can bring their own model key — your account, your cost, zero platform markup.
| Model | Best for | Cost | Billed via |
|---|---|---|---|
Claude Haiku 4.5 Budget | Diagnosis & summarisation | $1 / $5 per MTok | Platform pool |
Claude Sonnet 4.6 Premium | Routine code repair | $3 / $15 per MTok | Platform pool |
Claude Opus 4.8 Frontier | Complex / escalated repair | $5 / $25 per MTok | Platform pool |
Bring your own model BYO-only | High-volume, cost-efficient | Your key, your bill | Your API account |
Anthropic prompt caching (up to 90% off cached input) and batch discounts are applied automatically. A typical routine repair costs under $0.10.
Every safety guarantee here maps to real code in the Care engine. If we can't prove it in the audit log, we don't claim it.
Care never reads .env files, API keys, auth tokens, or payment credentials.
Schema migrations and payment table changes always wait for a human approval.
All autonomous code writes require your secure admin approval. Without it, every fix is downgraded to a reviewable preview — zero silent changes.
Platform-billed AI calls go exclusively to Anthropic — no third-party model ever touches your codebase on the platform pool.
Encrypted at rest and in transit. Built to respect your data and privacy obligations. Audit trails are tamper-evident.
Code-repair endpoints require your secure admin approval and are hardened against prompt-injection from external sources.
Add the telemetry snippet and Liulum Care starts watching. No migration, no downtime, no new infrastructure to manage.
Questions? Read the FAQ or read the full terms.