A PrimesFlow account takes about a minute to set up. Every new workspace starts on a 14-day free trial of the Pro plan — no credit card required.

1. Sign up. Create your account at primesflow.io/signup with an email and password.
2. Confirm your email. We send a verification link so we know the address is yours. The trial clock only starts once you've confirmed.
3. Create a workspace. A workspace is the container for your hosts, team members, alert rules, and integrations. Most companies keep one workspace; agencies typically create one workspace per client environment.
4. Install the agent on your first host — see the Quickstart below.

PrimesFlow installs with a single command. On any supported Linux host, run:

curl -sL primesflow.io/install.sh | bash

The installer downloads the agent, registers your host with your workspace, and starts an unprivileged primesflow-agent systemd service. Shortly after installation you'll see your first health signal in the dashboard, including CPU, memory, disk, and network baselines.

From there, add more hosts by re-running the same command on each machine — they'll appear automatically grouped under your workspace.

The PrimesFlow agent runs as an unprivileged systemd service and communicates outbound-only over HTTPS to api.primesflow.io. It does not open any inbound network ports and does not require ongoing root access after installation.

# install
curl -sL primesflow.io/install.sh | bash

# verify it's running
systemctl status primesflow-agent

The service runs as the primesflow user with least-privilege access to the metrics it needs. Logs are available via journalctl -u primesflow-agent.

Configuration lives at /etc/primesflow/agent.toml. Most teams never need to touch it, but you can tune the metric interval and label hosts to fit your environment.

# /etc/primesflow/agent.toml

[host]
name = "prod-api-01"
tags = ["production", "api", "hetzner"]

[metrics]
interval = "10s"   # 5s – 60s

After editing, reload the agent with systemctl restart primesflow-agent.

Each PrimesFlow agent reports a consistent set of infrastructure signals:

Collection intervals

• Hobby — every 5 minutes
• Pro — every 30 seconds
• Team — every 10 seconds

The dashboard gives you a per-host view with time-series charts and a fleet view that rolls every host in the workspace into one screen — grouped by tag so you can spot the one machine drifting from the pack.

When an incident opens, PrimesFlow drafts a plain-language summary that correlates the signals already collected by the agent: system metrics, service state, top process data, and recent deploy markers (if you've wired them in via webhook).

Each summary includes

• What happened, in one sentence
• When it started and how long it's been active
• Which hosts, services, or containers were affected
• Likely cause, ranked by supporting evidence
• Suggested next steps to investigate or mitigate

AI incident diagnostics are included on the Pro and Team plans. The Team plan adds deeper multi-host correlation across the fleet.

Data boundary: PrimesFlow analyzes infrastructure signals only — never your application code, user data, or file contents.

AI-generated insights are assistive and may be incomplete or incorrect. Review the available evidence and verify all actions before making production changes.

PrimesFlow can deliver alerts to Slack, Discord, email, or any HTTPS webhook. Add integrations from Settings → Integrations in the dashboard, then attach them to alert rules with per-channel severity filtering.

• Set thresholds for CPU, memory, disk, and custom service checks.
• Define quiet hours per channel to avoid paging humans overnight for warnings.
• Route critical incidents to a paging channel and warnings to a lower-priority room.

# example webhook payload
{
  "incident_id": "INC-2148",
  "severity": "critical",
  "host": "prod-api-01",
  "summary": "Memory leak in nginx-proxy",
  "opened_at": "2026-07-28T14:22:07Z"
}

PrimesFlow opens an incident when a metric crosses a configured threshold or a detector flags an anomaly — for example, steady RSS growth on a long-lived process. Each incident collects the metrics, events, and recent deploys around the window it fired in.

On top of that context, PrimesFlow drafts an AI incident summary with clear possible causes, a suggested next step, and links to the raw signals it used.

Invite teammates from Settings → Team. Each member gets an email invitation and joins the workspace with the role you assign.

Roles (RBAC)

• Owner — full access, billing, workspace deletion. One per workspace.
• Admin — manage hosts, alerts, integrations, and members.
• Member — view everything, acknowledge and resolve incidents, edit alert rules.
• Viewer — read-only access to dashboards and incidents.

Role-based access control is available on the Team plan. Hobby and Pro workspaces treat all members as admins.

Server groups

Group hosts by environment, client, or service (e.g. production, staging, client-acme). Groups scope both alert rules and dashboards, so a rule you create forproduction only fires on hosts tagged into that group.

Plan changes

• Upgrades take effect immediately and are prorated for the current billing period.
• Downgrades take effect at the start of the next billing cycle so you keep what you've paid for.
• Cancellation stops future billing; your workspace stays active until the end of the current period.

Payment

Payments are handled by a PCI DSS Level 1 certified payment provider. Card details are never stored on PrimesFlow servers.

Invoices

Invoices are emailed to the workspace Owner after each successful charge and are also available from Settings → Billing.

Data retention after cancellation

After you cancel, we retain your workspace data for 30 days so you can reactivate without losing history. After that window, metric history and incident records are permanently deleted.

Agent not connecting

1. Check the service is running: systemctl status primesflow-agent
2. Confirm outbound HTTPS to api.primesflow.io is not blocked by a firewall.
3. Verify the enrollment token in /etc/primesflow/agent.toml matches your workspace.
4. Read the agent logs: journalctl -u primesflow-agent -f

systemctl status primesflow-agent
journalctl -u primesflow-agent -f
curl -v https://api.primesflow.io/healthz

Metrics look stale

Stale metrics usually mean the agent is running but can't reach the API. Follow the connectivity steps above. If the host itself is under heavy load, samples can also be delayed — check the agent's own CPU usage in top.

Alerts not firing

• Confirm the rule is enabled.
• Check the duration — a threshold must hold for the configured window before firing.
• Send a test event from Settings → Integrations to confirm the channel works.
• Review the rule's cooldown — recently fired rules stay quiet until the cooldown elapses.

Agent using too much CPU or memory

The agent is designed to be lightweight, but on small VPS instances you can reduce its footprint:

• Increase the collection interval in agent.toml.
• Reduce the number of services under active check.
• Update to the latest agent — recent releases include performance fixes.

You're always in control. To fully remove the agent from a host, run:

curl -sL primesflow.io/uninstall.sh | bash

The script stops the systemd service, removes the agent binary and /etc/primesflow, and deregisters the host from your workspace.