Agents can help you design complete application features — forms, lists, dashboards, and more — using natural language. Describe what you need and the agent generates the configuration.
Part 1: Creating features
Start a conversation with the agent and describe the feature you want to build. For example:
- “Create a feature that shows a list of customers with their name, email, and status, and lets me click to see the detail”
- “Build a form to register new employees with fields for name, department, and hire date”
The agent will generate the feature and let you preview it before applying.
Feature type
Agents build custom features (React code) by default — they cover forms, lists and dashboards, plus any logic a JSON configuration cannot express, so the agent no longer asks you to pick a type.
Ask for the standard type explicitly if you want a JSON-configured feature instead:
- “Create a standard feature for customers”
- “Build the customer list as a JSON configuration feature”
Charts
Custom features can include charts. Ask for them the way you would describe any screen:
- “Create a dashboard with production orders by state as a bar chart and sales per month as a line chart”
- “Add a donut with the share of each product category to the sales feature”
The generated code uses the platform’s chart component (@codenull/charts), which follows your application’s theme colors and handles tooltips, legends and empty states. Charts take aggregated data (one row per category or date; up to 200 rows and 6 series), which the generated code requests from the backend with each table’s Aggregate query and a pie shows up to 8 slices, grouping the rest as “Otros”. For charts outside its scope (dual axes, scatter, radar) the agent can use recharts directly. The same modules are available when you build features through the MCP server.
Part 2: Refining and updating features
After the initial feature is created, you can ask the agent to refine it:
- “Add a filter by status to the customer list”
- “Change the form layout to two columns”
- “Add a button that navigates to the order history of the selected customer”
The agent understands the context of the existing feature and applies only the requested changes.
Be specific about field names and data sources when prompting the agent. The closer your description matches your data model, the more accurate the generated feature will be.