Agent Analytics
Server Details
Analytics your AI agent can actually use. Track, experiment, and optimize via MCP.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- OAuth
- Works in Glama
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 32 tools
Most tools are individually distinct, but several pairs blur together: properties vs properties_received have near-identical descriptions ('discover event names and property keys' vs 'property keys by event name'), sessions vs analytics_sessions differ only in record list vs histogram, and bot_traffic_overview vs all_sites_bot_traffic differ only by scope. The sprawling query tool also explicitly overlaps every analytics_* convenience wrapper, so an agent must guess whether a shortcut or the ad-hoc query is the intended path.
All names are snake_case, but conventions are mixed: an analytics_* prefixed family (overview, funnel, retention, pages...) coexists with bare nouns (query, errors, sessions, properties, compare, live_now) that are equally analytical, plus verb_noun CRUD for experiments/projects/annotations. The verb-based tools like append_project_annotation and web_vitals break the pattern further.
32 tools exceeds the 25+ threshold and is heavy for what is fundamentally a read-analytics server. Many entries (analytics_breakdown, analytics_heatmap, analytics_pages, analytics_insights, all_sites_overview) are thin presets that the flexible query tool already covers, so tool budget is spent on redundancy rather than capability.
Coverage is broad: project lifecycle (create/list), full experiment CRUD, project context get/set/append, billing, and a deep read surface spanning funnels, retention, paths, vitals, errors, and a powerful ad-hoc query. Minor gaps exist (no delete/update project, no annotation or context-entry removal, no per-user profile read), but core analytical workflows have no dead ends.
Available Tools
32 toolsaccount_usageARead-onlyIdempotentInspect
Read current account plan, usage, billing estimate and spend cap, Pro pricing, and a human payment handoff when needed. Requires account:read. Read-only: cannot charge or manage billing. Call again after checkout to verify hosted entitlement, then retry the original task.
Inputs: no named parameters.
all_sites_bot_trafficBRead-onlyInspect
Filtered automated traffic across all active projects in your account.
Inputs: limit, period.
all_sites_overviewARead-onlyInspect
Historical summary across all projects in your account: total events, active project count, daily trend, and top projects.
Inputs: limit, period.
analytics_breakdownBRead-onlyInspect
Top property values ranked by count (pages, referrers, UTM sources, etc.) Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required), days, event, limit, +1 more.
analytics_funnelARead-onlyInspect
Funnel analysis: track where users drop off across a sequence of events (e.g. page_view → signup → purchase). Returns per-step user counts, conversion rates, and average time between steps. Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: steps (required), project (required), since, count_by, +3 more.
analytics_heatmapARead-onlyInspect
Day-of-week × hour traffic grid with peak detection (times in UTC) Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required).
analytics_insightsARead-onlyInspect
Period-over-period comparison with trend (events, users, bounce rate, session duration) Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required), period.
analytics_overviewBRead-onlyInspect
Get time series chart, KPIs (events, users), and country breakdown (events, unique users) for a project over N days Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required), days.
analytics_pagesARead-onlyInspect
Entry/exit page stats with bounce rate, avg duration, events per session Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required), type, limit.
analytics_panelAnalytics explorerARead-onlyIdempotentInspect
Use this when you want to explore an authorized project's analytics beside the conversation. Opens a compact project/view picker; accepts empty input. Analysis queries retain their existing plan and access checks. Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: no named parameters.
analytics_pathsARead-onlyInspect
Bounded session path analysis from entry pages to goal, drop-off, or truncation. Useful for connecting page behavior to funnels, retention, and experiments. Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required), goal_event (required), since, max_steps, +3 more.
analytics_retentionARead-onlyInspect
Cohort retention analysis — track what % of users return over time. By default uses session-based retention (any return visit counts). Pass 'event' to switch to event-based retention. Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required), event, period, cohorts, +1 more.
analytics_sessionsBRead-onlyInspect
Session duration histogram with engaged percentage and median bucket Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required).
append_project_annotationAIdempotentInspect
Append one annotation to project context. Append-only: goals, activation events, glossary entries, and existing annotations stay in place. Allowed fields are project, occurred_at (ISO-8601 timestamp, stored as UTC), title (max 120 characters), and an optional note (max 500 characters). The annotation identity is normalized occurred_at plus title. A duplicate returns 200 with appended false and keeps the stored note. The 100 annotation cap returns 409 and stores nothing. Use this for a meaningful landing page, pricing, onboarding, feature, release, or experiment change. Do not store git commit logs, noisy edits, PII, secrets, or long notes.
Inputs: title (required), project (required), occurred_at (required), note.
bot_traffic_overviewARead-onlyInspect
Filtered automated traffic for a single project: automated requests, dropped events, categories, and top actors.
Inputs: project (required), limit, period.
compareARead-onlyInspect
Compare a shipped change and return a keep, revert, or extend verdict. Compare is a free read. Each successful call counts as one metered read. Pass ship_date plus window_days, or explicit baseline and compare windows. window_days is an integer from 1 to 90 and a multiple of 7 (default 7, so the largest ship window is 84). The ship day is in neither window: baseline is ship-N through ship-1, and compare is ship+1 through ship+N. Run a day-7 read on ship day + 8 so the last day is complete. If either window includes today UTC, the verdict is extend with reason incomplete_current_day, and that reason overrides every other verdict. goal_event is required and is counted as unique converters (count_mode unique_users): a user who fires the goal more than once counts once. denominator_event defaults to page_view and is counted as unique users. rate = goal_converters / denominator_unique_users. Optional filters use the query filter shape and apply as they do today. A filter on event is rejected. Optional denominator_filters use the same {field, op, value} shape and operators and apply only to denominator events when computing denominator_unique_users. goal_converters still counts only goal users inside that filtered denominator. Omit denominator_filters or pass [] for today's behavior. Max 10 entries. To exclude one-hit 404 page views from the denominator: denominator_filters [{"field":"properties.title","op":"neq","value":"Page not found | grokbotpages.com"}]. rule thresholds are relative fractions (0.15 means +15%). The absolute value of keep_threshold and revert_threshold must be at most 5. Omitted rule fields use the documented day-7 examples (keep_threshold 0.15, revert_threshold -0.1, min_denominator 200, min_goal_events 20 unique converters across both windows). Those examples are not product policy. Verdict order: extend incomplete_current_day; extend below_min_sample; extend zero_denominator; extend zero_baseline when lift is null; revert revert_threshold_met only when lift is at or below revert_threshold and the 90% interval lies entirely below zero; extend inconclusive when the drop is within noise; keep keep_threshold_met when lift is at or above keep_threshold and interval.low is above 0; extend interval_includes_zero; extend no_evidence_of_gain. A drop by itself is not a revert. Each window reports date_from, date_to, day_count, denominator_unique_users, goal_unique_users, goal_converters, goal_users_outside_denominator, count_mode, and rate. The response also includes count_mode, reason_detail, lift, echoed denominator_filters, and a 90% interval (level, method newcombe_wilson_90, estimate rate_difference, low, high, excludes_zero, note). Optional metric compares a numeric property. metric is sum(), avg(), p50(), p75(), or p95(), for example sum(revenue) or p75(cwv_lcp). The property uses the query property allowlist ^[A-Za-z0-9_]+$ with a maximum of 128 characters. Example: metric p75(cwv_lcp), metric_event page_view, better lower, while goal_event stays signup. On hosted, cwv_* vitals are properties on page_view. Prefer the web_vitals tool for the p75 headline, ratings, and shares. metric_event is the event whose property is measured and defaults to goal_event. Set it when the measured event is not the goal. better is higher (default) or lower and is not inferred from the property name. Use better=lower for cwv_lcp and the other Web Vitals. Shared filters apply to the metric event. denominator_filters still apply only to the denominator. sum() compares the sum divided by unique denominator users (per_user true). avg and percentiles compare the aggregate itself (per_user false). Percentiles are nearest-rank, so percentile_method is nearest_rank. With metric set, interval is null and interval_method is none. There is no bootstrap interval in v1. lift is (compare - baseline) / baseline, negated when better is lower, so the existing thresholds keep their meaning. raw_lift is the unnegated ratio. min_denominator is still unique denominator users per window. When metric is set, min_goal_events is numeric values (n) on metric_event across both windows, not unique converters. below_min_sample is extend when n across both windows is under min_goal_events (default 20) or either window is under min_denominator (default 200). incomplete_current_day, below_min_sample, zero_denominator, and zero_baseline stay in that order. A missing compare value, or a baseline of 0 or null, is zero_baseline with lift null. When lift meets keep_threshold or revert_threshold, the verdict is extend with reason no_interval_for_metric, and lift is still reported. Lift between the thresholds is extend / no_evidence_of_gain. The response metric object includes metric, fn, property, metric_event, better, per_user, and percentile_method. Each window still includes the conversion rate and user counts, plus value, n, and skipped. Omit metric and the compare response is unchanged, including the Newcombe interval. Compare stays a free read when metric is set.
Inputs: project (required), goal_event (required), rule, better, +9 more.
create_experimentCIdempotentInspect
Create an A/B experiment
Inputs: name (required), project (required), variants (required), goal_event (required), +1 more.
create_projectAIdempotentInspect
Create a new Agent Analytics project. Returns the tracking snippet and API example. Idempotent: returns the existing project if the name is already taken. The tracking snippet includes data-track-vitals="true" so new installs collect Core Web Vitals on page_view.
Inputs: name (required), allowed_origins.
delete_experimentCDestructiveIdempotentInspect
Delete an experiment
Inputs: id (required).
errorsARead-onlyInspect
Read grouped $error events for a project (GET /errors). Errors is a Pro read. A free account gets 403 PRO_REQUIRED, the same status and error code as POST /query, and that rejection is not metered. Each successful call counts as one metered read. Errors is the post-ship guardrail next to compare: after a ship, read errors for the same window before you keep the change. The group key is a SHA-256 fingerprint of the normalized message (URL query strings and hashes removed, then UUIDs, hex hashes, and numbers replaced), the source filename without query or hash, and the numeric line. There is no stack, and col is not part of the key. Each group includes fingerprint, a sample message, source, line, count, users, sessions, first_seen, last_seen, and up to 3 top_paths. count, users, sessions, and top_paths use the since/until window. first_seen and last_seen use every stored $error for that fingerprint. Groups with no event in the window are omitted. since defaults to 7 days and accepts 7d or YYYY-MM-DD. until defaults to today UTC. limit defaults to 20 and maxes at 100. new_since is a YYYY-MM-DD or an annotation title. A date is used directly. Any other string matches project annotations by exact title, then case-insensitive title, then note, then substring. Zero matches is 400. More than one match is 400 with matches listing occurred_at and title. The group new_since boolean is present only when this param is set, and is true when first_seen is on or after the resolved date. new_since_annotation is returned only when new_since matched an annotation. Every response includes approximate, a boolean. When approximate is true, caps is an array of the scan caps that were hit: distinct_triples and window_rows. There is no singular cap field. A large project stays 200. The response also includes truncated.
Inputs: project (required), limit, since, until, +1 more.
get_experimentBRead-onlyInspect
Get experiment details with results Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: id (required).
get_project_contextARead-onlyInspect
Get compact self-improving project context: goals, activation events, event-name glossary, and major date annotations used to interpret analytics results. Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required).
list_experimentsBRead-onlyInspect
List A/B experiments for a project Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required).
list_projectsARead-onlyInspect
List all your Agent Analytics projects Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: no named parameters.
live_nowARead-onlyInspect
Real-time snapshot: active visitors, sessions, events per minute, top pages, and recent events. Reads from in-memory ring buffer — no D1 query.
Inputs: project (required), window.
propertiesBRead-onlyInspect
Discover event names and property keys from recent events Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required), days.
properties_receivedCRead-onlyInspect
Property keys by event name, sampled from recent events Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required), since, sample.
queryARead-onlyInspect
Run a flexible ad-hoc analytics query with custom metrics, filters, grouping, date ranges, and privacy-first customer email lookup. This is the most powerful endpoint — use it when other tools don't answer the question.
Examples of questions this tool answers:
"How many signups from Germany this week?" → filters: [{field:"event",op:"eq",value:"signup"},{field:"country",op:"eq",value:"DE"}]
"Which events contain 'page' in the name?" → filters: [{field:"event",op:"contains",value:"page"}], group_by: ["event"]
"Daily unique users for the last 30 days" → metrics: ["unique_users"], group_by: ["date"], date_from: "30d"
"Events per country" → group_by: ["country"]
"Signups by path and day" → metrics: ["event_count"], filters: [{field:"event",op:"eq",value:"signup"}], group_by: ["date", "properties.path"]
"p75 of LCP" → use the web_vitals tool (p75 is the headline). For an ad-hoc percentile on hosted, metrics: ["p75"], property: "cwv_lcp", filters: [{field:"event",op:"eq",value:"page_view"}]
"Revenue collected" → metrics: ["sum"], property: "revenue"
Filter operators: eq, neq, gt, lt, gte, lte, contains Filterable fields: event, user_id, date, country, session_id, timestamp, and any properties.* field (e.g. properties.path) For customer-specific reads, use the top-level email input instead of hashing locally. Raw email is sent only in the authenticated HTTPS POST body; the server matches via a project-scoped HMAC index and does not store raw email in event rows or profile traits. Built-in fields are a closed list. Event properties such as referrer, utm_source, path, browser, and hostname must be queried as properties.referrer, properties.utm_source, properties.path, properties.browser, and properties.hostname. Invalid filter fields fail loudly and return /properties-style guidance instead of being silently ignored. Group by: event, date, user_id, session_id, country, and at most one properties.. The key matches ^[A-Za-z0-9_]{1,128}$, the same rule as property filters. Result rows use the column name properties.. A null or missing property value is grouped as null. Two properties.* entries return 400 INVALID_GROUP_BY. Metrics: event_count, unique_users, session_count, bounce_rate, avg_duration, sum, avg, p50, p75, p95 Numeric property metrics sum, avg, p50, p75, and p95 are Pro, with the same gating as this query. A free account gets 403 PRO_REQUIRED and that rejection is not metered. property is the event property key and is required when any of those metrics is requested. On hosted, cwv_lcp, cwv_inp, cwv_cls, cwv_fcp, and cwv_ttfb are properties on page_view when the tracker snippet includes data-track-vitals="true". Use the web_vitals tool for the p75 headline, ratings, and good/needs-improvement/poor shares. For an ad-hoc query, filter event to page_view. property must match ^[A-Za-z0-9_]+$ and be at most 128 characters. Missing property is 400 MISSING_FIELDS. An invalid key is 400 INVALID_PROPERTY_KEY. Only JSON numbers are aggregated. Numeric strings, booleans, nulls, and missing keys are skipped. Each row includes the requested metrics plus n (numeric values used) and skipped whenever an aggregation metric is requested. Percentiles are nearest-rank: percentile_method is nearest_rank when a percentile is requested and null for sum or avg only. avg is rounded to 6 decimal places. order_by may be a requested aggregate, for example p75. Count modes: raw, session_then_user. The default for event_count is session_then_user for activation-safe counting: session-backed rows count by session, no-session rows fall back to user only when that user has no session-backed row in the same group, and fully anonymous rows fall back to event id. count_mode is ignored when event_count is not requested.
Inputs: project (required), email, limit, order, +8 more.
sessionsBRead-onlyInspect
List individual session records Anonymous preview is limited to agentanalytics-demo synthetic read-only data; sign in for your own projects.
Inputs: project (required), limit, since.
set_project_contextADestructiveIdempotentInspect
Set or replace compact project context. The encoded API request body is capped at 512KB. Read existing context first, merge still-valid entries, inspect event names with properties or properties_received, and save only durable product truth tied to goals, activation events, event-name glossary entries, or date annotations for major product changes. Use annotations for meaningful landing page, pricing, onboarding, feature, release, or experiment changes; do not store git commit logs, noisy edits, PII, secrets, or long release notes.
Inputs: project (required), goals, glossary, annotations, +1 more.
update_experimentCDestructiveIdempotentInspect
Update experiment status (pause, resume, complete)
Inputs: id (required), status (required), winner.
web_vitalsARead-onlyInspect
Read Core Web Vitals for a project (POST /vitals). p75 is the headline: each metric is assessed at the 75th percentile of page loads, using nearest-rank. Web Vitals is a Pro read. A free account gets 403 PRO_REQUIRED, the same status and error code as POST /query, and that rejection is not metered. Each successful call counts as one metered read.
The event is fixed to page_view. On hosted, vitals are cwv_lcp, cwv_inp, cwv_cls, cwv_fcp, and cwv_ttfb properties on page_view. Collect them by installing the project snippet with data-track-vitals="true".
Thresholds follow web.dev (https://web.dev/articles/vitals) and PageSpeed Insights. LCP: good at or below 2500 ms, poor above 4000 ms. INP: good at or below 200 ms, poor above 500 ms. CLS: good at or below 0.1, poor above 0.25 (unitless). FCP: good at or below 1800 ms, poor above 3000 ms. TTFB: good at or below 800 ms, poor above 1800 ms. A value at or below good_max is good. A value above poor_min is poor. Any value in between is needs-improvement. The same rule rates every page load for the good, needs-improvement, and poor share, and rates the p75.
cwv_pass is true only when the LCP, INP, and CLS p75 ratings are all good. If INP n is 0, the decision uses LCP and CLS only. cwv_pass is null when LCP n is 0 or CLS n is 0. FCP and TTFB do not affect cwv_pass.
group_by is optional. device returns mobile, tablet, and desktop. path returns the top limit paths by sample size, sorted by worst LCP p75. Omit group_by for one site-wide row (group null). limit defaults to 10 and maxes at 20. days, date_from, and date_to select the window. Each metric row includes p75 with its unit, rating, n, skipped, the good/needs-improvement/poor share, and up to 5 top poor targets. FCP and TTFB targets are empty. skipped counts page views that did not report that vital. When the scan hits the 200,000 event cap, approximate is true.
Inputs: project (required), days, limit, date_to, +2 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
MCP observability. Query live traffic, errors, duration, and alerts from your AI agent.
Clamp Analytics MCP server: traffic, revenue, funnels, cohorts, errors, and search, for AI agents.
Analytics for MCP servers. Find out which of your tools agents get wrong. MCPulse shows you which tools AI agents retry, which come back empty, and which they never call at all. Two lines inside your own server. It never sees your arguments or your results. getmcpulse.com
MCP server for building and testing AI agents with multi-model experimentation and insights.
Related MCP Servers
- AlicenseAqualityDmaintenancePrivacy friendly, cookieless web analytics built MCP-first. "Add analytics to my Next.js app" → an AI agent runs the setup_analytics_for_site tool, picks the right install snippet, edits your layout file, and verifies the script is loading. OAuth onboarding, no API keys to paste.281MIT
- AlicenseAqualityDmaintenanceFirst-party web analytics MCP server for AI agents, providing 42 tools to query traffic, events, funnels, conversions, sources, and performance data.4045 npmMIT
- AlicenseBqualityBmaintenanceCookieless web analytics your AI agent can read. Self-host it on your own Cloudflare account (Workers + D1) and read visits, sources, pages, countries and custom events over MCP. Prefer hosted? https://a.measuremy.site/mcp (OAuth sign-in, no key).201MIT
- AlicenseNot gradedqualityDmaintenanceProvides unified AI agent observability including tracing, cost tracking, performance monitoring, anomaly detection, and audit trails via MCP.56 npmMIT
Glama MCP Gateway
Add one secure layer between your agents and this server.