Can the Metabase server build a funnel or retention cohort directly?

Not as native analytics primitives. It works with dashboards and cards over your connected databases, so a funnel would be expressed as SQL-backed cards rather than a built-in funnel tool. For native funnels and retention, PostHog or Mixpanel fit the job more directly.

What does this server actually let an agent do in Metabase?

Read and manage dashboards (list_dashboards, get_dashboard, create_dashboard), arrange the cards on them (add_card_to_dashboard, update_dashboard_card), run a card to get its result (execute_dashboard_card), and share a board with create_public_link, all over the databases connected to your instance.

Metabase for product analytics

Pick 3 of 4 for product analyticsCommunityCognition57

Product analytics is about answering behavioral questions, who converts, where users drop off, against the platform that holds your data. This maintained Metabase server is our third pick of four here, and its angle is the BI layer rather than the event pipeline: it lets an agent run questions and work with dashboards over the databases you have connected to Metabase.

It ranks third because Metabase is general business intelligence over SQL sources, not a purpose-built product-analytics engine. Where your behavioral data already lives in a warehouse and you reason about it with SQL through Metabase, this server fits. For event-native funnels and retention, the picks ahead of it are closer to the job.

How Metabase fits

The tools here center on dashboards. list_dashboards and get_dashboard read what exists, create_dashboard, update_dashboard, copy_dashboard, and delete_dashboard manage them, and the card tools, get_dashboard_cards, add_card_to_dashboard, remove_card_from_dashboard, and update_dashboard_card, compose the individual queries on a board. execute_dashboard_card runs a single card to pull its current result, and create_public_link shares a dashboard out.

The honest limit: this tool set is built around dashboards and their cards, so it is strongest for assembling and reading curated views rather than ad hoc one-off queries, and it inherits Metabase's SQL-over-connected-databases model rather than a native event schema. PostHog is the stronger pick for product analytics with built-in funnels, retention, and event capture. Mixpanel is purpose-built for behavioral event analysis. ClickHouse fits when you query raw event data directly at scale. Reach for Metabase when your analytics is self-serve BI across databases and dashboards are how your team consumes it.

Tools you would use

Tool	What it does
list_dashboards	List all dashboards in the instance.
get_dashboard	Get a dashboard's details.
create_dashboard	Create a new dashboard.
update_dashboard	Update an existing dashboard.
delete_dashboard	Delete a dashboard.
copy_dashboard	Copy an existing dashboard.
get_dashboard_cards	List the cards placed on a dashboard.
add_card_to_dashboard	Add a card to a dashboard.
remove_card_from_dashboard	Remove a card from a dashboard.
update_dashboard_card	Update a card's placement or settings on a dashboard.

Full Metabase setup and config →

FAQ

Can the Metabase server build a funnel or retention cohort directly?: Not as native analytics primitives. It works with dashboards and cards over your connected databases, so a funnel would be expressed as SQL-backed cards rather than a built-in funnel tool. For native funnels and retention, PostHog or Mixpanel fit the job more directly.
What does this server actually let an agent do in Metabase?: Read and manage dashboards (list_dashboards, get_dashboard, create_dashboard), arrange the cards on them (add_card_to_dashboard, update_dashboard_card), run a card to get its result (execute_dashboard_card), and share a board with create_public_link, all over the databases connected to your instance.