Studio Foreground

Platform

The Foreground Suite splits game authoring into the part that needs judgement and the part that needs precision. AI helps with interpretation. Ordinary, testable code does the precise work of writing engine files. The writer approves what moves between them.

How the Foreground Suite fits togetherForeground Lore scans the game project and gives its canon to Foreground Script. In Foreground Script the writer drafts with AI and approves, the scene becomes a structured description, and a plain-code compiler writes standard event commands into the project. Foreground Stage and Foreground Light save data files that runtime plugins play in the game. No AI runs in the finished game.Foreground LoreForeground ScriptForeground StageForeground LightScans your project:switches, variables,events and pluginconflictsMenus, HUDs,dialogue stages,battle screensLighting, shadows,weather andeffects for maps1Write with AI2You approve3Scene description4Compiler (plain code)Your RPG Maker MZ projectMaps and eventsEvent commandsPlugin data filesNo AI runs in the finished game.scancanonwritesdata files

How the parts connect

  • Foreground Lore scans the game project into a local database: switches, variables, events and how they connect.
  • Foreground Script uses that database as canon while you write. The approved scene becomes a structured description, and the compiler writes it into the project. Afterwards the scene is checked against the project and the project is rescanned, so the next scene starts from up-to-date canon.
  • Foreground Stage and Foreground Light are visual editors. They save their work as data files, and runtime plugins play those files in the game. Your events call them with standard plugin commands.

Principles

  • Deterministic last mile. The step that writes engine files is plain code. The same input gives the same output, and anything unrecognised is marked and reported, never guessed.
  • Human approval gates. AI proposes and the writer decides: which draft, which facts become canon, and when a scene is written.
  • Canon-aware context. The writing assistant works from your scanned project and approved notes, not from a generic idea of what your game might be.
  • An agent harness for creative work. Foreground Script uses the loop-and-check design of AI coding agents (retrieve context, act, verify), rebuilt for writing.
  • Provider-agnostic. Works with any provider that offers the standard OpenAI-compatible API, hosted or on your own machine, switched by configuration.
  • Nothing AI at runtime. The finished game contains standard engine data and plugins. Players need nothing extra.

Engineering

The Foreground Script pipelineFour steps: Write, Interpret, Compile, Verify. The writer approves between writing and interpretation, and again before anything is written to the project. Writing and interpretation are AI-assisted; compiling is plain code; verification is automatic checks.1WriteDraft with an AIthat knows yourcanon2InterpretProse becomes astructured scenedescription3CompileStandard eventcommands, writtenby plain code4VerifyChecked againstyour project, thenrescanned▶ Writer approves▶ Writer approvesAI-assistedPlain, testable codeAutomatic checksNothing reaches the game until the writer approves it. No AI runs in the finished game.

We build tools that write into people's game projects, so we test them hard:

  • Thousands of automated tests across the suite.
  • In-game verification. Key runtime features are checked inside the running game, including a control run that must fail, so a passing check means something.
  • Targeted mutation checks. On critical runtime logic, we deliberately introduce bugs to confirm the tests catch them.
  • Low-end hardware first. Visual features are measured on a dedicated low-end test machine with integrated and entry-level graphics.
In private development

All four products are proprietary software. The roadmap shows what works today.

See it in each product: