Building A Dashboard

A dashboard is a collection of insight tiles on one page. Your job is to figure out which insights belong on it, reuse what already exists, create what’s missing, and lay them out sensibly — not to blindly generate charts.

Create vs update

First work out whether you’re creating a new dashboard or changing an existing one.

• Search existing dashboards with dashboards-get-all (its search param does fuzzy name/description matching). If the user is clearly describing something that already exists, they probably want an update.
• Read a candidate with dashboard-get to see its current tiles before you change anything.
• If the request is ambiguous — “get my financial metrics together” could mean build new or add to an existing one — ask a short clarifying question rather than guessing.

Use templates as reference

PostHog ships vetted dashboard templates for common topics, and orgs can share their own. Consult them before you build — they’re a strong signal of which insights pair well on a topic.

1. dashboard-templates-list — browse templates (use search for a topic, scope to narrow to global / team / organization). This returns names, descriptions, and tags only.
2. dashboard-templates-retrieve — open the closest template to see its tiles: which insights it groups together and how each is queried.

Treat templates as examples, not a spec. Take inspiration from the insights and their groupings, but tailor every insight to the user’s own events, properties, and intent. Don’t copy a template verbatim, and don’t force a template onto a request it doesn’t fit — a good bespoke dashboard beats a mismatched template every time.

Select the insights

Prefer reusing existing insights over recreating them.

• Search with insights-list and read promising ones with insight-get to check they match the user’s intent and actually have data. Full-text search misses things named differently, so list broadly before concluding an insight doesn’t exist.
• For anything missing, create it with insight-create (see the product-analytics insight skills for query shape).
• Keep the set minimal — only the insights the request needs. A focused dashboard is more useful than an exhaustive one.

Assemble the dashboard

• New dashboard: dashboard-create with a short (3–7 word) name and a concise description, then add the insight tiles.
• Existing dashboard: dashboard-update. Adding, replacing, or removing insights means sending the full intended set of tiles — insights you omit are removed, so include the ones you want to keep.
• Layout: by default preserve existing tile placement. Use dashboard-update to plan each tile independently on the 12-column grid. Tile widths can be any whole number from 1 to 12, subject to each tile’s minimum size. Use wider tiles for primary charts and smaller tiles for supporting metrics. Mixed rows such as 8 plus 4 or 6 plus 6 can show that hierarchy.
• Reflow: use dashboard-reorder-tiles only when the user explicitly asks to reorder tiles or make every tile the same size. Its layout modes give every tile a uniform box. For mixed widths or heights, use dashboard-update.
• Tile sizes: send tiles through dashboard-update with each tile’s id and a complete layouts.sm box. sm is required whenever you send layouts, because a write replaces the tile’s whole layout. sm controls desktop placement, and the dashboard derives the mobile layout from the sm order and heights, so set only sm. The API stores only x, y, w, and h. It does not resolve overlaps, so plan the grid before you send it.
• Verify with dashboard-insights-run to confirm the tiles return data, then summarize what you built and invite the user to refine it.

When not to use this

• Saving a single insight — just create the insight; it doesn’t need a dashboard.
• Adding non-insight widget tiles (text cards, widgets) — see the widget tools (dashboard-widget-catalog-list, dashboard-widgets-batch-add) instead.

Related skills

• managing-subscriptions — deliver the finished dashboard to email or Slack on a schedule
• creating-ai-subscription — a recurring AI-written report, when prose beats a wall of charts