Meet maxinja

maxinja is the agent your team briefs inside maxclicks. Give it a customer outcome and the facts that define the message; it can investigate the data, build a campaign or journey, and refine the result. Generated conditions and calculations also appear in their own app controls.

Audience conditions, evaluated attributes and workflow logic use the same pattern: describe the intended behavior, inspect the explanation and results, then refine it. Your team directs that behavior in plain language.

Where you meet maxinja

maxinja shows up in two forms.

• The assistant pane. A conversation you open alongside your work. It answers platform questions, queries your space's real data with safe read-only tools, and builds broadcasts, email templates, and workflows end to end (name, audience, schedule, and content). Attach the page you are on with the Current view chip, or @-mention an entity (a contact, object, schema, segment, template, broadcast, workflow, webhook, form, or brand kit) to give it more context. One exchange runs at a time. Messages sent while maxinja is working steer that active work; include the correction and which part of the result it changes.
• Inline "Ask maxinja" prompts. The describe-what-you-want field on a page. This is how you generate a segment's filter, an evaluated attribute, and workflow logic.

Describe, review, refine

The logic maxinja writes runs on maxclicks servers. It never appears in the app and the API never returns it. What you review instead is a plain description of what it does. See evaluated attributes for where that logic gets used.

Ask for customers who signed up in the last 30 days but haven't ordered, and maxinja shows something like:

Contacts created within the last 30 days with no related Order records.

If that matches what you meant, confirm it. If not, refine (only count completed orders) and it regenerates. Three habits make this reliable:

• Describe the outcome and identify the source context it requires. Supply missing campaign facts rather than expecting maxinja to infer them.
• Review the description and representative results, including records that should be excluded.
• Be specific. rewrite this in a confident tone for a first-time buyer beats make this better.

Separate authoring from execution

The editor’s maxinja input can refine the current subject or block; it edits saved content. An Adaptyle prompt runs later as recipient content is rendered. A published workflow executes its own saved steps. These are different actions with different costs and states.

Workspace assistant runs require write access. State whether you want a draft or an execution action. Messages sent during work steer it at a checkpoint; stopping does not reverse completed edits or sends. Open the saved result and verify its state before continuing.

maxinja can also manage typed Track catalog definitions and background computed attributes, queue a calculation refresh and inspect published values. Define the source, subject and time window so the measure answers the intended question. This catalog work does not ingest observations or create source credentials; use the Events setup controls for a collected event's source, Properties and Context contract. A refresh request is not a completed publication.

What it costs

AI generation uses credits. Distinguish work done during setup from AI that runs for each recipient or workflow execution. Model, context, previews and regeneration all affect usage; use the billing report to inspect actual spend.

Next

maxinja reference: exchanges, interactions, and scoping →