agent srv_8f3a2c connected · awaiting instructions

Infraa manages your Linux server. You just approve.

The AI agent that installs services, watches for problems, and fixes things on any Linux VPS. Connect a server in 30 seconds. Then chat your way through everything else.

perplesmile@laptop ~
$ curl install.infraa.sh | sh
→ downloading agent
→ verifying signature ✓
→ registering with gateway…
✓ connected as srv_8f3a2c
ready in 28s
$
srv_8f3a2c · prod-api-01
live
Install Postgres 16 with nightly backups
infraa
I'll propose 4 steps. You approve before anything runs.
1Install postgresql-16 (apt)
2Configure postgresql.conf + tune memory
3Schedule nightly pg_dump backups
4Run first backup + verify freshness
ApproveReview commands
Approved · run it
infraa
Done in 47s · 4 actions logged.
connect
~30s
agent
single static binary
health checks
every 60s
writes
signed · single-use
01How it works

Three things, done well — install, chat, watch.

01

One-line install

Run one curl command on any Ubuntu, Debian, or Rocky Linux VPS. The agent registers and starts watching.

$ curl install.infraa.sh | sh
downloading agent
verifying signature
registering with gateway
connected as srv_8f3a2c
Works onubuntudebianarm64amd64ready in 28s
02core

Chat to operate

The AI proposes a plan. You approve, Infraa runs it. Every action is logged and auditable.

Install Postgres 16 with nightly backups
infraa
Proposing 4 steps. Approve to run.
1Install postgresql-16 (apt)
2Configure postgresql.conf + tune memory
3Schedule nightly pg_dump backups
4Run first backup + verify freshness
ApproveReview commands
Approved · run it
infraa
Done in 47s · 4 actions logged.
apt install postgresql-162.3s
postgresql.conf written0.1s
nightly pg_dump scheduled1.2s
first backup verified43.4s
03

Watches while you sleep

Tracks disk, memory, CPU, and services. Flags trouble in your dashboard before things break.

srv_8f3a2cchecks every 60s
CPU
14%
Memory
2.1 / 8 GB
Disk
62 / 80 GB
disk77.5%
recommendation
/var/log growing 1.2 GB/day. Rotate?
03App deploys

Your code ships here too.

The same server that runs your database can run your app. Import a GitHub repository and every git push builds in your own GitHub Actions, ships as a checksummed artifact, and goes live behind nginx with free HTTPS.

Eight framework presets, zero YAML. You approve every release — or turn on auto-deploy and forget it.

04Security model

Built for people who don't want to hand over root access.

Infraa doesn't get root. The agent runs as a non-root user with a narrow allowlist of commands. Every state-changing action requires your cryptographically signed approval.

The AI proposes — you approve — the agent executes. Defense in depth, fully auditable.

If you're not comfortable with what's about to happen, you can read the exact commands before they run. Built to earn trust, not assume it.

01
LLM proposes
Read-only tools. Drafts a plan. Never executes.
02
You approve
One-tap review in chat. Inspect raw commands.
signed token · 5 min · single use
03 · agent executes
Verifies signature, arg-hash, expiry. Runs only the approved capability. Logged, auditable.
Capabilities
code-reviewed
Modules
data only
Tokens
RS256 / 5m
05Bring your own AI

Already pay for Claude? Point it at your servers.

Infraa is an MCP server. Connect the agent you already use — one command, one OAuth sign-in — and it can see your servers, read live metrics and logs, and draft fixes. The thinking runs on the AI subscription you already have.

An outside agent gets exactly the guardrails Infraa's own AI gets: reads are free-flowing, every write becomes a proposal that waits for your signed approval in the dashboard. No exceptions, whichever model is talking.

Works withClaude Codeclaude.aiCodex CLIGemini CLICursor
claude — ~/work
mcp
> why is prod-api-01 slow right now?
read_metrics(server_id: srv_8f3a2c)
list_processes(top_n: 10)
CPU is at 94% — a Postgres autovacuum has been stuck on events_logfor 40 minutes. I've proposed a fix:
propose_capability(service.restart, postgresql)
proposal prop_4e12 waiting for your approvalinfraa.sh/app
06FAQ

Questions, answered before you ask.

The short version: the AI proposes, you approve, a locked-down agent executes. The longer versions are on the right.

01

What is Infraa?

Infraa is an AI server management platform for Linux VPSes. You connect a server with one curl command, then install services, monitor health, and fix problems through chat. The AI proposes each action, you approve it, and a locked-down agent on your server executes it — no SSH session required.

02

How does Infraa manage a server without SSH?

A small agent runs on your server and holds an outbound connection to Infraa — no inbound ports to open, no SSH keys to share. When you ask for something in chat, the AI drafts a plan; once you approve it, the agent runs only the approved, cryptographically signed commands.

03

Is it safe to let an AI manage my server?

The AI in Infraa never executes anything directly. Every state-changing action requires your explicit approval, delivered to the agent as a single-use signed token that expires in five minutes. The agent itself runs as a non-root user with a fixed, code-reviewed allowlist of capabilities.

04

Which Linux distributions does Infraa support?

Infraa supports Ubuntu 22.04, Ubuntu 24.04, and Debian 12 on both amd64 and arm64, on any VPS or cloud provider — Hetzner, DigitalOcean, AWS EC2, Linode, and others. Installation is a single curl one-liner that registers the server in about 30 seconds.

05

What services can Infraa install?

Infraa installs and manages services like Postgres, MariaDB, Redis, nginx with Let's Encrypt certificates, and Docker from a one-click catalog or a chat message. After installing, it keeps monitoring each service's health and flags problems in your dashboard before they become outages.

06

How is Infraa different from Coolify or Dokploy?

Coolify and Dokploy deploy applications onto a server that is already set up. Infraa is the layer before: it sets up and manages the server itself — installs the services, watches them, and fixes them — through AI chat. You can use Infraa alone or underneath an app-deployment platform.

07

Can I use my own Claude or ChatGPT subscription with Infraa?

Yes. Infraa ships an MCP server, so any MCP-capable agent — Claude Code, claude.ai, Codex CLI, Gemini CLI, or Cursor — connects with one command and operates your servers on the AI subscription you already pay for. Connected agents read live state and propose changes, but every write still requires your signed approval in the Infraa dashboard. The BYO AI plan starts at $12/month.

08

How much does Infraa cost?

Infraa plans start at $12/month for BYO AI, where your own Claude subscription drives Infraa over MCP; hosted-AI plans are Standard at $25/month and Pro at $50/month, and annual billing saves 20%. Every plan includes the agent, the chat-first dashboard, one-click service installs, and monitoring — there is no per-server billing.