Whagons assistant

The Whagons assistant is an in-product collaborator that works in the signed-in user's tenant context. It can explain, navigate, inspect, analyze, export, and—when authorized—apply changes through curated product actions.

It is not the same as an external agent calling the REST API.

What the assistant understands

The assistant has compact, product-maintained skill documentation for:

• workspaces, tasks, filters, exports, and comments;
• users, teams, roles, settings, and personal preferences;
• task types, templates, forms, custom fields, statuses, transitions, priorities, tags, and SLAs;
• approvals, acknowledgments, workflows, KPI cards, reminders, and notifications;
• integrations and API keys;
• CSV and Excel inspection, analysis, and imports;
• messages and workspace chat;
• Workbench Write documents, Grid spreadsheets, and Present decks;
• product routes and safe navigation.

Read, preview, apply

The assistant separates work into three modes:

1. Read — inspect visible state, reference data, task sets, settings, files, or workbench content.
2. Preview — validate mappings, imports, forms, templates, or structured edits before mutation.
3. Apply — run an approved curated action and verify the result.

This pattern is especially important for bulk imports, destructive changes, permission edits, and document operations.

Files and data

The assistant can inspect attached CSV, XLSX, and XLS files, page through large datasets, profile columns, query data, map source fields to Whagons concepts, preview skipped values, and apply supported imports.

For large tenant task sets, it can create a permission-filtered data snapshot for analysis rather than loading every row into chat context.

Navigation and exports

The assistant can open safe in-product routes, prepare a filtered task view, and export task results to supported formats such as PDF, Excel, CSV, or JSON.

Product writes

Writes use named, curated actions with validated payloads. The current families are listed in Assistant capabilities.

The available result always depends on current user permissions, tenant configuration, enabled powerups, and backend rollout state.

What the assistant will not do

• It will not bypass Whagons permissions.
• It will not claim a write succeeded without a confirmed tool result.
• It should not ask users to paste secrets into chat.
• It should not run a destructive action without clear intent and required confirmation.
• It should not invent a task, workspace, user, field, or action when live data is required.

Read Safety and permissions for the full user-facing model.