Unity · React Native · Capacitor · your backend
Analytics for your game that you just ask questions of.
Give your agent the skill, it integrates the SDK. Then ask it where players quit, who comes back, what they pay for, and whether the last patch broke the economy. No taxonomy to design, no dashboard to build, no BI hire.
Free plan: 1 million events a month, every feature, no credit card.
$ ah funnel tutorial_start level_1_complete first_purchase --by props.platformFUNNEL 3 steps — by props.platform (18,402 entered)s1 tutorial_start · s2 level_1_complete · s3 first_purchase props.platform s1 s2 s3 conv%ios 9,614 7,210 418 4.3%android 8,788 5,102 161 1.8% TIME TO CONVERT (within-session, all dims)transition n median p90s1 → s2 12,312 3m 40s 11m 05ss2 → s3 579 6m 12s 19m 30s $ ah changes add "Android: skip tutorial step 3" --kind experimentlogged — tomorrow's digest shows it against the funnel Just ask
Questions, not dashboards
Every answer is a terminal table, or --json for a script. You ask from a shell; your coding agent asks exactly the same way.
$ ah funnel tutorial_start tutorial_step_3 level_1_complete --since 7dFUNNEL tutorial_start → tutorial_step_3 → level_1_complete — 18,402 sessions entered # step sessions %entered drop1 tutorial_start 18,402 100% —2 tutorial_step_3 11,930 65% -35%3 level_1_complete 9,874 54% -17% TIME TO CONVERT (within-session)transition n median p901 → 2 11,930 1m 12s 4m 05s2 → 3 9,874 48s 2m 30sSDKs
Drop it into the engine you already ship
One project key across every platform and your backend — every player lands in the same place, and a question about iOS vs Android is one flag.
Unity
A UPM package. Scene loads and uGUI taps are captured for you; gameplay events are one call with any props you like.
AgentHog.Init(new AgentHogConfig {
Host = "http://agenthog.io",
ProjectKey = "ah_xxxxxxxx",
});
AgentHog.Capture("level_complete",
new Dictionary<string, object> { ["level"] = 12 });React Native / Expo
A provider at the root. Screen views and taps autocaptured, persistent ids via AsyncStorage.
<AgentHogProvider config={{
host: 'http://agenthog.io',
projectKey: 'ah_xxxxxxxx',
appName: 'dungeon-dash',
}}>
{children}
</AgentHogProvider>Capacitor
One init call for a hybrid game. DOM autocapture like the web tracker, native transport underneath.
import { AgentHog } from '@brightmotion/agenthog-capacitor'
await AgentHog.init({
host: 'http://agenthog.io',
projectKey: 'ah_xxxxxxxx',
})Server-side ingest
Relay App Store, Play, RevenueCat, or Stripe webhooks and backend events with a write token — the purchases a client never sees.
curl -s http://agenthog.io/ingest \
-H "Authorization: Bearer $AGENTHOG_INGEST_TOKEN" \
-d '{"project":"ah_xxxxxxxx","anonId":"server:42",
"events":[{"name":"subscription_started",
"props":{"tier":"plus","price_usd":9.99}}]}'What you get
Everything a studio asks about — nothing to design first
Send events with the names you already use. Funnels, cohorts, revenue, and economy meaning are applied afterwards, and apply retroactively to everything you already sent.
Funnels & retention
Any sequence of events is a funnel, split by platform, version, country, or any prop. Cohorts by day, week, or month.
Aug 17's cohort is holding at 46% in week one — up from 39% two weeks ago.
$ ah retention --cohort week --periods 4
Revenue — real money
Map the amount prop on the purchase event you already send, relay store webhooks server-side, and get net revenue, ARPPU, and LTV by SKU, platform, or variant. Retroactive, refund-aware, never summed across currencies.
net (30d)
$18,340
payers
811
ARPPU
$22.61
starter_packgems_1200season_passremove_ads$ ah revenue --by props.sku --since 30d
Economy monitoring
Declare which events are sources and sinks of which currency — after the fact, no SDK change. See supply drift, paid vs earned, the faucets and drains that dominate, and whether a patch broke any of it.
$ ah economy map --event gems_purchased --source && ah economy
Feature flags & experiments
Deterministic variants evaluated in the SDK, exposure tracked for you, Bayesian results. Difficulty curves, shop layouts, tutorial length — tested on real players.
tutorial_skip_v2running · 9dmetric: level_1_completecontrol · 6,410 exposed56.0%skip · 6,388 exposed62.5%98.7% chance to beat control+11.6% lift — ship it, then retire the flag.
$ ah experiments results tutorial_skip_v2
Every patch, lined up with the metric it moved
Log builds, live-ops beats, and balance changes to the changelog — backdatable — then read D1 retention or ARPPU against them. When a number moves, there's a suspect.
$ ah changes add "Patch 1.4.2 — easier level 3" --kind deploy --at 2026-08-12
Bot & test-device filtering
Every session is scored human, bot, or crawler, and your own QA devices read as test. Store review bots and your build farm never inflate D1.
Raw SQL and export — own your data
Read-only SQL over your own events, on every plan. Pull everything out as JSON whenever you like; nothing is locked behind a tier.
Built for your coding agent
The ah CLI is the whole read surface — no MCP server, no plugin. Hand an agent a read-scoped token and it can answer the questions above itself.
Why not the usual tools
Built for the studio without a BI person
Game analytics suites hand you dashboards if your game fits their taxonomy. Product-analytics tools hand you a blank canvas and a tracking plan to write. Neither is a question you can just ask.
| Game analytics suites | Web product analytics | AgentHog | |
|---|---|---|---|
| Before you can ask anything | Fit your game into a fixed taxonomy — design, progression, resource, business events. | Write a tracking plan up front, then build the dashboards that answer each question. | Send events with whatever names you already use. Meaning is applied later — and retroactively. |
| Getting an answer | Open the prebuilt dashboard, if there is one for the question. | Find the dashboard someone built, or build a new one. | One command from a shell or your coding agent. --json for scripts. |
| Game-shaped reads | Progression and economy dashboards, out of the box. | Product analytics shaped for web and SaaS; games are a stretch. | Funnels by platform, retention cohorts, revenue by SKU, experiments, and virtual-currency economy monitoring. |
| Your raw events | Raw export is a paid add-on. | Available, gated by tier or self-hosting. | Read-only SQL and JSON export on every plan, including free. |
| Free plan | Free tier with feature and export limits. | Usage-based; the useful features climb the tiers. | 1M events a month, every feature, no credit card. |
Think GameAnalytics on the left and PostHog or Mixpanel in the middle. Our reading of their public docs as of August 2026 — check theirs; they change.
How it works
Ship the SDK tonight, ask your first question tomorrow
Three steps, and none of them is a taxonomy workshop.
Create a free account
Sign up, create a project, and you get one key that every platform and your backend share.
# one project, every platform
ah_xxxxxxxxAdd the SDK
One init call. Scenes, screens, and taps are captured for you; gameplay and shop events are one line each, named however you like.
AgentHog.Init(new AgentHogConfig {
Host = "http://agenthog.io", ProjectKey = "ah_xxxxxxxx" });
AgentHog.Capture("boss_defeated",
new Dictionary<string, object> { ["boss"] = "Crypt Lord" });Ask
Install the CLI, log in, and ask. Your coding agent uses the same commands — every one takes --json.
npm i -g @brightmotion/agenthog
ah login && ah digest
ah funnel tutorial_start level_1_completeOr hand the whole thing to your coding agent
One command installs the AgentHog skill into Claude Code. Then ask it to instrument your Unity project — it knows where the init call goes, how to name events so the CLI can read them, and how to verify events actually arrived.
Install the agent skillFree for your first million events a month.
Every feature — funnels, retention, revenue, experiments, every SDK, raw SQL — is in the free plan. Pro is $49 a month when you want 12 months of history and more projects.
No credit card. Your events are yours to export, on any plan.
Up to 3 teammates
On the free plan; 20 on Pro. Never per seat.
30 days of history
12 months on Pro — same features either way.
Full CLI & API
For you and your agent, on every plan.
Retroactive by design
Map revenue or economy later; it applies to everything you already sent.