Ask your game why players quit.

Investigate player behavior, compare releases and record what you will try next. Review evidence from comparable games and public sources, then turn selected findings into marketing and publisher drafts. Every source stays traceable, and your game facts remain separate from external claims.

One autoload script, four fields in the Inspector, done. It handles device IDs, session tokens, batches of up to 50 events per call, an offline queue with exponential backoff, and a server-side kill switch you can flip without shipping an update.

Ravensight.track_event("level_completed", {
    "level": 3,
    "score": 15000,
    "time_seconds": 87.4
})

Rage-quit, stuck, and engaged players are flagged as the data lands. D1 and D7 retention cohorts and level completion funnels are computed server-side, so there is nothing to configure first.

See the routes players actually take through your game, including the loops they get caught in. The dashboard ranks where players stop, which transitions bleed players and where they bounce back and forth, with the raw transitions underneath. Ask the analyst about them, or query them over MCP.

Background digests summarize recent play and suggest where to look next. Generated work uses the shared AI allowance and prepaid balance.

Connect your AI client to scoped game analytics and studio findings with a personal access token. Each token carries its own monthly usage budget, set by the owner, and draws from your prepaid balance.

Name an ordered sequence of two to eight events and see how many reached each step, how many were lost against the step before, and the average and median time it took them. Count players or sessions, filter by an event property, and ask the analyst the same funnel across every game in the studio at once.

Connect GitHub and every push scans what changed, or upload a source ZIP or submit from CI. Compare what your code reports with live telemetry, with file references behind each finding. Get a quote before a scan; later scans focus on changed files, and read-only access is all we ask for.

Write your design docs here and talk to an analyst that already knows your studio. It reads your findings across every game you have shipped, proposes edits to the document in front of you, and leaves the writing to you: nothing changes until you accept it. Describe a plan and it will tell you which of your own past findings argue against it.

Add the optional Godot report overlay so players and QA testers can capture a screenshot, highlight an issue and leave a message. Triage reports, read the session events that led up to one, group them into issues and verify the fix.

Run the Ravensight Playtest CLI against your own build, on your own machine. It hands your game to personas with different ways of playing and brings back a report each: what worked, what they got stuck on, and whether they would have kept going, with findings you can mark real, intended or duplicate. Mark one intended and the next run stops reporting it.

Your source and your build never leave your machine, and the personas' sessions land in your own event stream tagged synthetic, excluded from every number unless you ask for them.

A project is one title. Its production, staging and development environments are separate games underneath, each with its own ingest key and its own event stream, so a debugging session cannot move your retention. Pick the project, then pick the environment.

Break any top-level key your events carry down by value: ranked values with event and player counts, an "other" bucket and a daily series. Then segment the player list by the same key to read the sessions behind a number.

Record the ads, videos, streams, posts and sales you run outside the game, by hand or from a CSV export, and the platform measures the new players in the 72 hours after each one against what a normal week would have brought anyway. It is a correlation with a stated method, not proof, and touchpoints without a fair comparison say so plainly instead of showing a made-up multiple. They show up as markers on your player charts too.

A board of cards beside a partner that has read your findings, your confirmed patterns, what your touchpoints did and what your titles earned before it answers. Every card can cite the evidence behind it, checked against your own studio, and the partner adds cards only to the board it is sitting at. Start blank, or fill a board from what the studio already knows for one AI unit, then promote the cards worth keeping into a design brief.

Record sales per storefront, period and currency by hand, or import any store export through a column mapping you confirm before a row is stored. Totals stay per currency with no converted figure, a monthly row is never added to a daily one, and a title with a release date gets its first ninety days as a launch curve. Recorded numbers only: nothing is read from a store on your behalf.

Invite your team by email as an admin or a member. Everyone shares one allowance and one bill, billing and member management stay with the owner, each person mints their own access token for their own agents, and the audit log records who did what.

Your events are yours to query. Write raw SQL against your own event store, read-only and scoped to your game, whenever you want the number yourself.

select level, count() as deaths
from events
where event = 'player_death'
group by level order by deaths desc