Orchestrator · Real-time data

Never ask a dev to redeploy a campaign again.

Wire your events once, then build unlimited campaigns, and reuse the same events across all of them. The AI orchestrator validates, enriches with live data, and handles the rest.

Orchestrator
Notifizz AI20/20 credits
› listLinkRoutes ·
› validateOrchestratorCode
The viewTaskUrl route also expects opportunityId, I'll add it from the event.
› saveOrchestratorCode
Review passed
Saved: review fully green. The “View task” button is now wired to viewTaskUrl with taskId and opportunityId from the event. You can promote the campaign.
Describe what your orchestrator should do…
Preview</> View code
Summary
When a task is created, fetch its detail and the author’s profile, then notify all platform users. The “View task” button links to the task via viewTaskUrl.
How it works
Event
The « task.created » event is received
Fires with userId, taskId and opportunityId.
Validation
opportunityId, userId & taskId are present
If one is missing, the orchestrator stops here: nothing is sent.
Enrichment · real-time
Fetch data via task.detail & user.profile
Live business data: nothing stored on Notifizz.
Dev tasksSend to review
No other task
DONE · 2
Done
Connect the Notifizz SDK
Done
Send a first event
Plug your event

Any event. Few tasks. Your LLM does the rest.

Send the events you already emit, even incomplete. The orchestrator handles almost everything: and the few cases it can't are turned into precise dev tasks your favourite LLM picks up and implements, with full context.

Plug it incomplete
The orchestrator reads the schema and flags any gap.
Your LLM is the interface
Whatever the orchestrator can't fill becomes a task your LLM picks up, with full context.
Works with your tools
Claude, Cursor, any MCP client. Nothing new to learn.
signal-task-oncreateDeclared
Property schema
userIddeclared + seenstring
taskIddeclared + seenstring
opportunityIdmissingstring
The orchestrator can't fill this, it raises a dev task
Add opportunityId to the event
Pending
Handed off to your own tool: outside Notifizz
Your LLM
Picks it up with full context · takes over
via MCP
Works with
ClaudeClaude CodeCursorAny MCP
ResolvedDone · no redeploy
Real-time data · Sovereignty

Enriched live. Your data, never seen.

At send time, the orchestrator calls your systems for exactly what it needs: uses it once, then drops it. No copy of your customer base, ever, and the AI never sees your users' data.

No copy of your customer base · AI doesn't see it · sovereignty
Developer experience

The best dev experience is the one you don't need.

Once the event is wired, every new campaign is built and shipped from the dashboard, no redeploy, no stack to babysit. Engineering stays in control; the team keeps moving.

1×
wire the event
0
redeploys after
∞
campaigns & reuse
Versioned campaigns
Every change ships as a version; in-flight sends finish on theirs.
Health monitoring
Each campaign reports healthy / degraded / broken, you know before users do.
Sequences & conditions
Chain steps with delays and "Only if…" rules: send the second only if the first went unread.
Transparent

See exactly what the AI does.

Nothing is a black box: inspect every execution, and understand what the AI generates before it ships.

Preview · before it ships

Understand what the AI generates.

Before anything goes live, the orchestrator shows a plain-language summary plus a step-by-step "how it works": event, enrichment, recipients, send. You know exactly what will happen.

Preview · Notification</> View code
Summary
On every Ghost post published, fetch all staff members (up to 100) and send them a notification-center message and an email about the new article.
How it works
Event
«ghost.post.published» is received
With post_id, url, primary_author.
Enrichment
Fetch data via Staff
Live: nothing stored.
Recipients
All staff members identified
Up to 100 per call.
Send
Notification sent
Channel: notification center + email.
Inspector · after it runs

Review every execution.

Each channel is validated, with the workflow run, the orchestrator trace, the spam score, the recipients and the step-by-step results: exactly what you review.

Workflow execution
✉ New article publishedSentSpam risk 3.9
Orchestrator trace
Staff · 91ms› PARAMS {…} · → RESULT {…}
Step results
Step 1 · completed232ms · 1 recipient
Step 2 · completed155ms · 1 recipient
Orchestrator FAQ

Questions about the orchestrator

Do I need to redeploy for a new campaign?
No: that is the whole point. Engineering wires an event once; every campaign after that is built and shipped from the dashboard. New campaign, new copy, new channels: zero redeploys.
What if my event is incomplete?
The orchestrator validates it and stops before anything is sent. What is missing becomes a precise dev task: resolvable from your IDE (Claude, Cursor…) with full context, via MCP. Nothing broken ever reaches a user.
Where does the live data go after sending?
Enrichers fetch it from your systems at send time, it is used once, then dropped. No copy of your customer base, and it is never used to train models.

Plug your first event today.

Let the AI orchestrate the rest: validation, live data, dev tasks and delivery.