Continual MI
MDL
Endless VN Engine
GamesMGPT APIEarningsMonitoringDocs
Learn
  • Introduction
  • Quick Start
  • MDL CLI
  • How It Works
Build
  • Workspace
  • Agentic Workflow
  • Instructions
  • Schema
  • Assets
Ship
  • Publishing
  • Monitoring
  • Earnings & Payouts
Reference
  • Package Model
  • API Keys & Auth
  • CLI Reference
  • Platform Reference
Docs/Monitoring

Monitoring

Track how players interact with your game through focused operational views for usage, generations, users, universes, and purchases. Monitoring is visible in navigation, but data sections require a Continual Pro creator account with hosted games.

Access model

Monitoring sections stay visible so creators know where operational data will live. If the signed-in account does not have Continual Pro or has no hosted games, each section shows a requirement panel instead of hiding the route.

Monitoring views

ViewWhat it shows
UsageCredit debits, request tokens, provider cost, billable cost, the active billing multiplier, and usage breakdowns by game or model.
GenerationsUnified text + special-scene image requests with filters for text-only, image-only, or both. Includes request metadata, response payloads, provider details, compare tabs, and image previews/reasoning where available.
UsersPlayers tied to your hosted games, including sign-in-only players.
UniversesSaved games — browse every saved step, inspect the full snapshot JSON, download JSON or text exports.
PurchasesPer-attempt traces spanning checkout creation, post-checkout confirmation, and Stripe webhook processing.

All views can be filtered to a specific hosted game, but they each land on the same focused surface.

Generation timing (MDL vs MGPT)

For MGPT-backed story steps, diagnostics split ownership explicitly. The server waterfall shows MDL-owned work: auth, access checks, the HTTP call to MGPT labeled mgpt_backend_roundtrip (wrapper time on the MDL host), parse/apply, response build, and background persist/usage/log flush. A separate MGPT section shows MGPT-owned stages returned inline by the model backend (storage load, hosted prompt assembly, message windowing, validation, and inner provider HTTP). The MDL wrapper is always larger than or equal to inner provider time because it includes network and MGPT server work. Rows without MGPT inline diagnostics show mgpt.status = missing: only the wrapper remains reliable; do not treat MDL stage names as a substitute for MGPT internals. RPC or transport failures before a valid model payload are recorded as mgpt.status = error with an error message (distinct from missing, which still indicates a successful HTTP round trip without inline MGPT diagnostics).

Understanding universes

A universe is a player's save file. Each universe tracks the current story state, full turn history, and runtime context. Players can have multiple universes (save slots).

The state count tells you how much a player has engaged — 50 states means about 50 turns played in that playthrough.

What to look for

Healthy signals

  • Players with multiple universes — they're replaying or experimenting.
  • High turn counts — players are going deep into the story.
  • Clean generation traces — requests succeed without unusual retries or asset failures.

Warning signs

  • Many short universes — players start but abandon early. Check your opening and instructions.
  • Repeated generation failures — prompt, asset, or infrastructure issues may be blocking play.
  • Stalls at similar turn counts — something in the story may be breaking at that point.
  • Repeated purchase failures — checkout, confirmation, or webhook processing may be broken.

Access

You only see monitoring data for games you own. Player data is private to the game creator, and non-creator or non-Pro accounts see the Continual Pro / hosted-game requirement state for each section.

Hosted games only

Monitoring is available for hosted games backed by Postgres. It currently does not show individual player transcripts, detailed funnel analytics, or real-time sessions.

Next

Ship
Publishing
Get players before you monitor them.
Build
Workspace guide
Return to building and refining.