Agents

The crew that runs this platform after launch.

Agents we build and run for you — not a slogan. Each one has a name, a job, and a rule for when it stops and asks a human. The four below run this platform live today; catalog agents show their own production status on each card.

Dispatch

Lead scoring & routing

What it watches
  • Contact form + chat-handoff submissions
  • Cal.com booking webhooks
  • Lead score changes
What it's allowed to do
  • Score a new lead and write it to the CRM instantly
  • Enroll a lead in the welcome or quote-followup email sequence
  • Send a hot-lead SMS alert to the founder at score ≥ 40
Escalation rule

A lead scoring ≥ 40, or any reply/booking, exits automation and waits for a human reply — never auto-closes a deal.

Scribe

Content & lifecycle ops

What it watches
  • Due sequence steps and scheduled broadcasts (15-minute cron)
  • Resource-magnet gate submissions
  • Every route needing an OG image
What it's allowed to do
  • Send the right email sequence step, in the subscriber’s own locale
  • Deliver a gated lead magnet the moment a submission comes in
  • Generate a fresh, on-brand OG image for any page on demand
Escalation rule

A send failure surfaces on the dashboard’s needs-attention queue; the compliance footer (unsubscribe + address) is appended by the engine itself and can’t be edited away, even by a human.

Sentinel

Integration health & uptime

What it watches
  • D1, Resend, Turnstile, Cal.com webhook, Twilio/SMS.ir — live status
  • Login rate-limits and session integrity
  • The 15-minute cron heartbeat itself
What it's allowed to do
  • Run a real, authenticated probe against every integration on demand
  • Rate-limit repeated failed logins per IP
  • Report exactly which integration is down — never guess "all good" without checking
Escalation rule

A failed integration check surfaces in /admin/settings as "not configured" or "failing" — never auto-retried silently, and never auto-fixed. A human decides the next step.

Analyst

Event tracking & KPI computation

What it watches
  • The full 15-event first-party taxonomy (page views, CTA clicks, conversions)
  • Email open/click tracking
  • Funnel and per-market KPI splits
What it's allowed to do
  • Insert every tracked event to the database, fire-and-forget
  • Compute traffic, funnel, and campaign stats on request for /admin/analytics
  • Split every KPI by market (EN/FA) so neither market’s numbers hide inside the other’s
Escalation rule

Read-only by design — Analyst never changes a lead, a send, or a subscriber. It only surfaces numbers a human then acts on.

Agents you can deploy on your own site.

The crew above runs Orbital itself. These are the same kind of named, escalation-bound agents, sold individually, running on any site — yours included, whether or not Orbital built it.

Live feed

Real, sanitized events from this platform’s own event log — no fake data.

  • Analyst · Recorded a page view

    14m ago
  • Analyst · Recorded a page view

    14m ago
  • Analyst · Recorded a page view

    1h ago
  • Analyst · Recorded a page view

    1h ago
  • Analyst · Recorded a page view

    1h ago
  • Analyst · Recorded a page view

    1h ago
  • Analyst · Recorded a page view

    2h ago
  • Analyst · Recorded a page view

    2h ago
  • Analyst · Recorded a page view

    2h ago
  • Analyst · Recorded a page view

    2h ago
  • Analyst · Recorded a page view

    3h ago
  • Analyst · Recorded a page view

    3h ago
  • Analyst · Recorded a page view

    3h ago
  • Analyst · Recorded a page view

    4h ago

How escalation works

Honesty builds trust: every agent has a threshold where it stops and a human decides, not the other way around.

  • Dispatch

    A lead scoring ≥ 40, or any reply/booking, exits automation and waits for a human reply — never auto-closes a deal.

    A human — founder or team
  • Scribe

    A send failure surfaces on the dashboard’s needs-attention queue; the compliance footer (unsubscribe + address) is appended by the engine itself and can’t be edited away, even by a human.

    A human — founder or team
  • Sentinel

    A failed integration check surfaces in /admin/settings as "not configured" or "failing" — never auto-retried silently, and never auto-fixed. A human decides the next step.

    A human — founder or team
  • Analyst

    Read-only by design — Analyst never changes a lead, a send, or a subscriber. It only surfaces numbers a human then acts on.

    A human — founder or team

Frequently asked

Three bands: task agents run $190–490/mo, department agents $790–1,900/mo, and autonomous operators $2,500–7,500/mo — each with a one-time setup fee. Enterprise and bespoke work is quoted per engagement.

Get a crew like this on your platform.

The Ops retainer puts a named agent behind every recurring task on your site, with the same escalation rules shown above.