Quickstart
From zero to a running agent in about 90 seconds.
Before you start
You need three things:
- A Linux host (x64 or arm64) with outbound HTTPS to
https://cp.agentina.digcub-lab.comon port 443. - An activation token — a one-time string that starts with
act_. - The trusted-keys file URL or path your operator gave you. The agent verifies license signatures against this set.
1. Install the agent
One command. The installer downloads the signed binary, verifies the cosign signature, and registers it under systemd.
agentina system user and installs to /opt/agentina/<version>/. It refuses any tarball whose cosign signature doesn't match a production key — so a MITM'd install fails fast.2. Verify it's running
From the command line
Expect activated: true, a non-empty agent_id, and a license.health: ok line.
From the customer portal
Sign in at /portal/login and open the Agents page. Your new agent should appear within 60 seconds with status active.
3. Confirm end-to-end
Watch the heartbeat land in journalctl:
You'll see a line like heartbeat: active → active every 60 seconds. That's the agent reporting to the Control Plane and getting back "keep going." You're live.
4. Where to go next
- Run your first real piece of work — concrete walk-through.
- Pick the right agent type — decision guide.
- Scale to a fleet — when one agent isn't enough.
- Glossary — every term in one place.
If something goes wrong
"invalid activation token"
Tokens are single-use and expire after 14 days. The first machine to redeem claims the token; if you ran the installer twice, the second run fails here. Mint a fresh one from /portal/tokens.
"failed to reach control plane"
The agent only initiates outbound HTTPS. Check egress to https://cp.agentina.digcub-lab.com on port 443. Corporate proxies generally need an explicit allow-list entry; set HTTPS_PROXY in /etc/agentina/agent.env if you go through one.
Anything else
Full troubleshooting checklist (logs, common errors, how to file a ticket) lives at /docs/troubleshooting.