Charts¶

The MCP Server’s Visualisation Studio lets the AI build reusable chart definitions straight from your Odoo data, preview them, embed them into Shared Pages, and tweak them over time. The chart engine renders with Apache ECharts under the hood and supports 9 chart types out of the box.

Where to find them¶

MCP Server ‣ Charts. Each row is an mcp.chart.ts record.

How charts are created¶

There are two paths:

AI-created — the user asks Claude / ChatGPT / Cursor “show me revenue by salesperson last quarter”. The AI calls generate_chart which inserts an mcp.chart.ts row. The Chart form displays the originating connection in the Created by AI Client field.
Hand-authored — an admin opens MCP Server ‣ Charts ‣ New, fills in the form, and saves. The Created by AI Client field stays empty.

The Chart form¶

Smart buttons¶

Preview — opens the rendered chart in a new tab. Disabled when no data source is set or the chart is archived.
Active / Inactive — flips the chart’s availability to compose_chart and Shared Pages. Archived charts stay in the database for audit.

Visualisation¶

Chart Type — bar, line, pie, scatter, gauge, radar, funnel, treemap, or sunburst.
Theme — command (dark) or light.

What to chart¶

The chart needs to know what to draw. Two ways:

Business Topic (preferred) — pick from the Ontology Concepts (Customer, Order, Invoice, Lead…). The same chart definition works across tenants because the ontology mapping picks the right Odoo model per company.
Specific Odoo Table (advanced) — pick a raw model name (e.g. helpdesk.ticket, mrp.production). Use this only when no Business Topic fits.

A guidance banner appears under this group while no source is picked — it walks new users through their first chart.

Aggregation¶

Group By — the field used to bucket records (e.g. country_id, date_order:month). Empty = a single KPI bucket.
Measure — the numeric field to aggregate (e.g. amount_total). Empty + aggregate=*count* = record counts.
Measure Aggregate — count, sum, avg, min, max.
Domain — an Odoo domain filter (the editor widget is pre-bound to the chosen target model).
Limit — max number of buckets to draw.
Order by — pick a field and direction, or write a multi-field sort by hand.

ECharts options¶

Under the ECharts Options notebook tab you can paste arbitrary Apache ECharts JSON to override the auto-built configuration (custom titles, colours, tooltips, legends…). The renderer shallow-merges your overrides on top of the default option. The default styling is intentionally polished — leave the field empty unless you have a specific reason to customise.

Description¶

Free-text note about what the chart shows, who it’s for, and anything a future editor should know.

The chart tools¶

generate_chart: Creates a new chart definition. Returns chart_id plus a clarifying ask_user prompt. The AI always asks the user what to do next — render now, share as URL, or both.
update_chart: Patches an existing chart. The AI calls this when the user iterates (“make it a line chart instead”, “filter to last 90 days”).
compose_chart: Renders the chart as HTML + a structured data_points payload for inline display. After rendering, the AI asks whether the user wants a public share URL.
save_chart_as_page: Persists the chart as a token-gated public URL anyone with the link can open. See Shared Pages.