Live server and domain metrics, audit findings, incident timelines, and remediation recipes — accessible to your team and to Claude Code, Cursor, and Windsurf via native MCP. One agent enrolls in seconds and catalogues every service, container, certificate, cron job and domain it finds — no asset registry, no YAML, nothing to declare. The fleet writes itself; you monitor, recover, and audit it from one fast console.
Most monitors stop at CPU · RAM · disk. We don't. The agent walks the host once and the rest of your fleet lights up — domains from web-server config, services from systemd, containers from Docker, certs, cron jobs, log paths — all monitored from the same install.
Dense, honest, and fast. No dashboards pretending to be features. Every pixel on screen does work.
One table for every server across every provider. Health, cost, region, uptime — all one glance.
View servers →Open a session on your laptop, close it on your phone. Multi-tab, split panes, key-synced, end-to-end encrypted.
Learn more →CPU, memory, disk, network — sparklines where they belong, full charts when you need them. No upsell.
See metrics →Ask why keywest spiked at 14:32. Get a real answer with log excerpts, not a paragraph of filler.
Try AI →Visual fleet topology with real-time pulse. Click a node — SSH, metrics, logs, all right there.
Explore map →Track 200 domains like 2. Renewals, DNS health, SSL expiry — with auto-renew and real Slack alerts that aren't noise.
Manage domains →curl | bash to a catalogued fleet, in seconds.Seven stations between the install command and the host being usable in your console — one per second, roughly.
postgres lock-wait p95 climbed 4× in the last hour on db-prod-01. why?
Two queries in pg_stat_statements account for 78 % of lock_wait_seconds in the last hour. Both run UPDATE orders SET … filtered by customer_id, which has no index.
CREATE INDEX CONCURRENTLY orders_customer_id ON orders (customer_id); ANALYZE orders;
Every assistant answer cites the metric, log line or config it grounded on. No hand-wavy paragraphs. Suggestions arrive as runnable recipes you can dry-run, audit, and roll back — not as advice you have to translate into commands.
pg_stat_statements / MySQL slow-log → suggested index, vacuum or config change. The query plan and the source rows are linked from the answer. Read-only by default. Write actions need an explicit scope. Per-server toggle hides hosts from the AI gateway entirely — and the toggle itself stays web-auth-only.
40+ commands covering the operator essentials — fleet status, server detail, service restart, recipes, logs, alerts, backups, attack mode, AI triage. Coloured tables, auto-width, ANSI-safe.
Do I need to install an agent?+
Yes, for the full feature set — discovery, auto-recovery, off-site backup, telemetry. Single binary, Apache 2.0, one-line install. Outbound-only; no inbound port required.
Does it work with my provider?+
Any Linux / FreeBSD / macOS host the agent can reach. First-party provider integrations today: Hetzner Cloud, DigitalOcean, Vultr, Linode, Cloudflare DNS, Hetzner Robot. AWS / GCP / Azure are on the roadmap; until then the agent works on those hosts, the cloud-side sync doesn't.
Where are my SSH keys stored?+
Encrypted in your OS keychain (macOS Keychain, Windows Credential Manager, libsecret). HostAtlas never reads them out except at connect time. SSH Gatekeeper additionally supports short-lived CA-signed certs so static keys aren’t required for gated access.
Is there a self-hosted version?+
Yes, on Enterprise — for air-gapped or sovereign deployments. The agent is always self-hosted; only the platform consumer side is a choice between our managed instance and your own.
How is this different from Datadog?+
Datadog watches. HostAtlas acts. We run auto-recovery before paging the human, attach pre-fire logs to incidents automatically, ship a CLI + MCP server that match the UI 1:1, sign the audit log. Pricing is per-host, not per-metric. The two coexist if you want them to.
What does it cost?+
Free tier covers a single host forever — no trial timer, no credit card. Starter covers 10 hosts, Pro 50, Enterprise unlimited. Off-site backup quota + AI features + retention scale with the tier. See pricing →
Same platform, different first ten minutes. Pick whichever sounds like a Tuesday you’ve had.
Auto-recovery before paging. Pre-fire logs auto-attached. Signed timeline you can actually defend at audit.
Read → For solo foundersSensible defaults. Encrypted off-site backup with a ransomware gate. AI triage when something goes sideways at 03:42.
Read → For developersCLI-first, MCP-native, open-source agent. Recipes, webhooks and ⌘K everywhere. No platform-engineering tax.
Read →Free tier on your first host, forever. No trial timer, no credit card. Install the agent and watch your fleet light up.