Niches · Product analytics & experimentation

Answer a count or trend question from tracked events

No company's docs show a full fit yet. The listings below do the job with a limit, or not at all.

0companies, full fit
13companies, fit with a limit
19companies carry the tag
24share of the category's work
13first-hand accounts

What counts as done

A number or series for the same event, filters and period as the vendor's own UI, with the definition and date range stated.

Measured, for our own tests: it matches the UI within 1 per cent, in at most 5 calls and 60 seconds. No docs we read state these figures, so they don't count against a listing here.

How agents ask for it: how many people used X last week, how is the pricing page doing since yesterday's deploy, why did revenue drop, what's our DAU, top pages by device

Does the job, with a limit 14

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

ListingWhy
Amplitude BB
Amplitude, Inc.
query_amplitude_data runs segmentation or queries a saved chart by id, and get_amplitude_context returns the project time zone and AI Context definitions, but large charts may be truncated with no flag and nothing tests UI parity. source
Countly E
Countly Ltd
drill_query computes counts, uniques and breakdowns with cursor paging, and events_summary covers event totals, but drill is an Enterprise plugin, period and timezone semantics aren't documented in the tools, and nothing tests UI parity. source
Fullstory B
Fullstory, Inc.
build_metric turns a description into a metric and compute_metric runs it with a time range and a UI link, and get_metric reads saved ones, but the MCP caps group-bys at 50 rows and the agent's translated definition may differ from a dashboard's. source
GrowthBook BB
GrowthBook, Inc.
POST metric and fact-table explorations run a saved fact metric over a lookback or custom date range against the team's warehouse, with a truncated flag, but date-boundary semantics aren't stated and nothing tests UI parity. source
LaunchDarkly BB
Catamorphic, Co. (dba LaunchDarkly)
search-aggregations counts Sessions, Metrics or Events with group-bys and buckets over the observability data, but groupByLimit caps groups, it covers only events sent to LaunchDarkly observability, and nothing ties it to a product analytics dashboard. source
Matomo B
InnoCraft Limited
Processed reports come from the same archives as the UI, in the site timezone, with inclusive ranges, but filter_limit returns only 100 rows unless set to -1, and incomplete periods aren't marked (acc-41, open). source
Mixpanel B
Mixpanel, Inc.
Run-Query runs insights queries and Get-Report reads a saved report with results, and Get-Business-Context supplies definitions, but Run-Query can't use Saved Metrics or cohort filters, sessions spend context on discovery (acc-20), and the cap is 600 requests an hour. source
Mostly Good Metrics E
Mostly Good Metrics
execute_query and query_metrics take a documented query definition (metric, filters, date_range presets or custom ranges) and comparisons use completed periods, but sketch-based uniques can return unavailable, and end-date inclusivity and timezone aren't stated. source
OpenPanel B
OpenPanel AB
get_report_data runs a saved report and aggregate tools cover overview, active users and top pages, but tools named pageviews counted different things (acc-33, closed 2026-10-27) and single-day queries returned zero (acc-32, closed 2026-10-17); neither fix was verified here. source
Pendo E
Pendo.io, Inc.
entityUsage, aggregateEntityUsage and appUsage give counts and time series over up to 367 days with historical or current metadata, and aggregate rows return hasMore with offset paging, but questions must map to Pendo Pages, Features and Track Events, and UI parity isn't tested. source
PostHog B
PostHog Inc. (formerly Hiberly Inc.)
query-trends and saved insights use PostHog's own query engine, and metric-list and data-catalog-metric-run expose governed metrics, but a SQL count returned the default limit plus one (acc-10, open) and HogQL can differ from the dashboard (acc-01). source
Statsig BB
Amplitude, Inc.
Dashboard widget results and metric values can be read over the Console API, but widget results use dashboard defaults with global filters and variables not applied, so numbers can differ from what the person sees. source
Userpilot B
Userpilot, Inc.
The MCP reads reports, dashboards, events and analytics results, but the docs list no tool names, parameters, date semantics or row limits; only the vendor's description backs it. source
Woopra D
Woopra, Inc.
/trends loads a saved trends report by report_id over a start and end in the project timezone, with limit, offset and a snapshot ID for stable paging past 1,000 rows, but only saved reports can be read, so a new question needs a report built in the UI first. source

Doesn't do this job 5

Listings in the category that do something else, with the reason.

ListingWhy
Flagsmith B
Bullet Train Ltd
Flagsmith is a flag and experiment service; it doesn't answer count or trend questions over tracked product events. get_feature_evaluation_data covers flag evaluations only. source
Harness Feature Management & Experimentation BB
Harness Inc.
FME measures experiment metrics from tracked events; the MCP and docs offer no count or trend query over events, only event-type listing. source
Heap D
Content Square, Inc.
Heap's developer APIs are write-only (track, identify, properties, deletion); no query, report or MCP endpoint is documented. source
Optimizely Experimentation B
Optimizely North America Inc.
Optimizely Experimentation answers experiment questions; exp_execute_query queries experimentation entities, not counts or trends over product events. source
RudderStack MCP C
RudderStack, Inc.
Tools report pipeline metrics (source event volume, delivery success, latency), not product questions over tracked events; analysis happens in the downstream warehouse or tool. source

What stops agents finishing

  • the agent writes its own query and its number differs from the dashboard
  • a row limit silently caps an aggregate
  • an exclusive end date returns zero for a single day
  • an incomplete period compared with a complete one

Around it

Before: code.merged-pr, observability.error-spike. After: messaging.post-summary, project-management.create-ticket.

Where this comes from

Weight 24 of 100 in Product analytics & experimentation: 13 of 55 account-to-use-case links (47 accounts, some in two use cases); the most described job, and where most wrong-number reports sit; counts scaled to 100 and rounded by hand

The same as data: /api/v1/use-cases/product-analytics.metric-question.json, or the MCP tool get_use_case with id product-analytics.metric-question. Other jobs in this category: /api/v1/niches?category=product-analytics.

Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person. A mark says what the docs show, not what a run measured.

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.