Custom Analytics is now live in StackAI. Admins get clear visibility into how every AI project is performing, through a dashboard they build and customize themselves. Create the widgets you want, choose the chart, metric and filter for each, and arrange them your way. Runs, latency, errors, token spend and adoption sit in one view, so it's obvious which agents are working and which need attention.

Key Takeaways

• Build the custom dashboards your team needs and get clear visibility into how every AI project is performing.

• Create widgets from runs, token usage, node and project data, with six chart types and custom filters.

• See exactly what's working, what's failing and where to improve next, backed by real usage data.

Running AI Agents Without Clear Visibility

Teams running AI agents in production face a steady stream of questions: which agent fails most often, where token spend is concentrated, and why a key workflow has started to slow down. Answering them quickly is what keeps AI projects reliable and on budget.

Until now, the analytics page offered the same fixed views to every user. Questions outside those views required manual work, such as exporting run logs to spreadsheets, rebuilding charts, or submitting a request to StackAI and waiting. By the time the numbers arrived, they were often out of date.

The larger cost was limited visibility. A failing agent could go unnoticed until users reported problems, and rising token spend could surface on an invoice before it surfaced anywhere else.

Adding more fixed metrics would only answer one question at a time, leaving the next one open.

Custom Analytics takes a different approach. It gives admins the tools to build their own views and get the answers they need to make confident business decisions.

Build a Dashboard in Five Steps

Custom Analytics lives on the Analytics page in StackAI, and a first dashboard takes only a few minutes to set up.

Create the dashboard. Open Analytics from the sidebar, select New dashboard under Dashboards, and give it a name and an optional description.

1. Add a widget. Select Add Widget and choose a chart type: stat, line, area, bar, pie or table.

2. Choose the data. Pick a dataset (Runs, Token Usage, Node Runs or Projects), then the metrics to show. Group results by project, model, provider or node type, and add filters such as a single project.

3. Label and size it. Add a title, an optional tooltip description and a width. A live preview shows the chart before it is saved.

4. Read and refine. Set the date range to update every widget at once, drag widgets into place, and edit any widget from its menu.

A single operations dashboard can combine total workflow runs, latency for a key agent, token consumption over time, node execution volume by type, projects by published status and token usage by model, all on one screen.

The same builder is available on each project's Analytics tab, where dashboards count only that project's data.

Start Building Your First Dashboard

Pick the report you rebuild every week and make it a dashboard. Then give each project team a view of its own. As your program grows, the dashboards grow with it, one widget at a time.

Want to see Custom Analytics? Get a demo with our team here.