Feature highlight

Describe the Dashboard You Want. Phaide Builds It.

A Brief Becomes a Working Dashboard

Phaide's Dashboard feature is an AI agent that turns a plain-language brief into a complete, multi-widget dashboard. You supply two things — a name and a purpose — and the agent writes the analysis code for every widget, runs it against your organization's data, and arranges the results on a responsive twelve-column grid.

Three widget types cover most reporting needs: KPI tiles for single headline numbers, interactive Plotly charts for trends and breakdowns, and data tables for detail. Each widget is real code executed against your governed dataset, not a static mock-up. The agent works in iterations — drafting a widget, running it, reading the result, and correcting course — so a finished dashboard is one that has already been tested end to end.

When a dashboard renders, every tile reads from the same snapshot of your data, so a KPI, a chart, and a table on the same screen always agree with one another. Every dashboard lives under Dashboards in the left navigation, alongside Chat, Explore, Knowledge, and Database, so the whole team works from one place and one source of truth.

No SQL, No Chart Config, No Waiting on an Analyst

The point of the feature is to remove the gap between a question and a chart. You describe the outcome you want; the agent handles the querying, the calculation, and the visualization.

It does this carefully. Before saving a dashboard, the agent runs each widget's code in a sandbox and confirms it produces a valid result. Tiles that would error never reach you — what you open is a dashboard that already renders.

Editing stays in natural language. Open the Chat panel beside any dashboard, mention a specific widget with an @ reference, and ask for a change — a different metric, a new time window, another chart style. The agent regenerates just that piece and the grid refreshes in place. Need an additional chart? The New button drops a fresh-chart request straight into the chat. Each dashboard keeps its own chat threads, so you can run separate lines of editing — one for the revenue tiles, another for the funnel chart — and return to either later without losing context.

Dashboards are built for teams. Sharing is role-based — owner, editor, and viewer — and you can also set an organization-wide default role so everyone gets appropriate access at once. Viewers can still open the chat to ask questions about a dashboard read-only, without any risk of changing it. Layout is yours to shape: drag tiles to reposition them and drag their edges to resize, and the arrangement saves automatically.

Because every widget is live code, dashboards stay current. Results are cached for instant loads, then recomputed against your latest ingested data. When the underlying data structure changes in a way a dashboard can no longer reflect, Phaide flags it and offers a one-click regenerate rather than showing you stale numbers.

Governance carries through the whole feature. Widget code reads your data in read-only mode, and results pass through the same masking that protects sensitive fields everywhere else in Phaide — so building a dashboard never widens who can see what. Access is decided by the role you were granted, and every action a dashboard offers, from viewing to sharing to deleting, is checked against it. The result is a self-serve reporting layer that stays inside the boundaries your organization already set.

From Blank Canvas to Built Dashboard in Four Steps

1. Open Dashboards. Select Dashboards from the left sidebar. If you have none yet, an empty state invites you to create your first; otherwise the New dashboard button sits at the top right of the list.

2. Name it and state its purpose. A dialog asks for a Name (for example, "Sales Dashboard") and a Purpose (for example, "Monthly sales KPIs and trends"). The name is required; the purpose guides what the agent builds. Click Create.

3. Let the agent generate. Phaide builds in the background, writing and testing each widget as it goes. A live progress view shows the steps it takes, and the finished dashboard opens automatically when it completes.

4. Refine and share. Rearrange tiles by dragging and resizing. Open Chat to edit conversationally, or use a widget's New option to add a chart. From the ⋮ actions menu, choose Share to assign roles and set the org-wide default, or Archive when a dashboard has run its course — archived dashboards can be restored later.

Your Next Dashboard Is One Sentence Away

The dashboards your team keeps meaning to build no longer need a ticket, a spare afternoon, or a free analyst. Write down what you want to see, and Phaide turns it into something live, shared, and ready before the meeting starts. Open Dashboards and describe the first one.

FAQ

Do I need to know SQL or Python to build a dashboard? No. You write a name and a purpose in plain language; the agent generates and validates all of the underlying code for you.

What kinds of widgets can a dashboard contain? Three types: KPI tiles for single headline numbers, interactive Plotly charts, and data tables. The agent chooses the right mix for your stated purpose.

Can I change a dashboard after it's built? Yes. Open the Chat panel to edit any widget conversationally, mention a specific tile with an @ reference, add new charts with the New button, and drag or resize tiles to adjust the layout.

Can I control who sees a dashboard? Yes. Sharing is role-based — owner, editor, and viewer — and you can also set an organization-wide default role so everyone receives appropriate access at once.

Can viewers make changes? No. Viewers have read-only access to the dashboard, but they can still open the chat to ask questions about it without modifying anything.

Do dashboards update when my data changes? Yes. Widgets recompute against your latest ingested data, with results cached for fast loads. If the data's structure changes so a dashboard can no longer be displayed, Phaide flags it and offers a one-click regenerate.

Where do I find my dashboards? Under Dashboards in the left navigation. The list shows every dashboard you can access, with an option to show archived ones.

Related
GuideHow to Use Phaide via MCP8 min readPerspectiveHow Do You Safely Connect an AI to a Production Database? Why a Read-Only User Isn't Enough10 min readAnnouncementPhaide AI is live: AI agents that explore your data and find problems for you4 min read

Give it a shot — this is something different.

Connect the MCP endpoint and ask your agent a real question about your data.

Try Free