One game creation studio. Room to grow.
Plan, build visually, create assets, write code, and playtest in one project. We’re building a conversational path for beginners, with the control to keep growing as a developer.
Start with the idea and a simple scene. Add detail as the game takes shape, keeping your project and your work together.
Choose a game profile or playable blank. Keep a project plan, or review and import one from Markdown.
Use simple placeholders to lay out your scene and try the gameplay while the visual direction develops.
Import or generate assets, review them, then replace supported blockout appearances while preserving identity and gameplay bindings.
Refine scenes and behavior with visual tools and scripts. Conversation is becoming another way to work on that same project.
Play the active scene, save and reopen your work, and export through the supported project and asset routes.
Planning and blockout replacement have passed focused development gates. The complete conversational beginner journey is still being integrated and verified. Check the current beta scope.
Our goal: let a beginner make a real game conversationally, then grow professionally inside that same project. Supported beginner journeys should require no code, while code remains a first-class way to create.
Start with guided choices, visual editing, and reviewable changes. Conversational authoring is in development, with durable plans that survive beyond a chat session.
Edit scripts, use visual graphs, inspect changes, and work from the integrated terminal. Keep the convenience of guided tools as your skills and project grow.
A dated snapshot of editing, Play, persistence, recovery, and export tested together. New development is checked separately as it joins the studio.
Historical internal acceptance results for the unified-creation Phase 1 checkpoint. These counts do not certify the current development build or compare performance with other engines. Read the evidence summary.
August 18, 2026 — unified-creation checkpoint: 344 unit/integration files; 3,950 tests passed and 9 skipped. Browser gate: 131 passed, 10 skipped, zero failed, one worker and zero retries. The checkpoint also records a passing production build, save/reload and recovery coverage, and four visible export checks comparing archive entries by hash.
September 8, 2026 — accepted development slices: blockout authoring and appearance replacement; durable project plans; reviewable Markdown import; and progress checks based on project evidence. These are bounded implementation results. Live conversational wiring and the uninterrupted beginner journey require further integration and verification.
Release scope: Chromium web beta. Packaged desktop, other browsers and individual export adapters need their own certification. This website update did not rerun or recertify the engine release gate.
Specialist workspaces share the same project, assets, scenes, jobs, history, and health context.
Project-aware help for creation and scripts. Durable planning is implemented; deeper conversational authoring and live planning context are still being connected.
Assistance today · broader authoring in developmentRun the authoritative active scene with input, cameras, interactions, HUD state, and sandboxed gameplay scripts.
Scene-aware Play lifecyclePortable project and simulation ZIPs, asset and map formats, and model output. Target-engine adapters vary in maturity and remain explicitly labeled.
Receipts and artifact hashesImport, voxel, sculpt, material, animation, physics, and export tools connected to the same project resource system.
Specialist workspaceKeep your project plan across sessions, review an imported plan, and distinguish proposed work from changes supported by project evidence. Save and recovery preserve the work itself.
One project through every workspaceGenerate or import, process, validate, review, retry, accept, place, reimport, version, and roll back durable assets.
Raw results remain candidatesMonaco editing, a sandboxed QuickJS terminal, visual graphs, debugging, and project-aware script iteration.
Traditional code stays first-classMap Studio, asset generation, ForgeWeld, and Power Studio can stand alone from the main menu while sharing the unified lifecycle when a project is open.
Preflight supported locations, capture attributable elevation data, review the exact terrain candidate, then create an undoable scene.
Managed USGS-derived DEM route in active validationPaint square, hex, and isometric terrain, use procedural generation, and keep an offline path when providers are unavailable.
Local-first terrain toolsUse compatible ComfyUI, procedural, hosted, BYOK, and imported routes without exposing raw provider plumbing to beginners.
Requested vs. resolved receiptsPreview processed artifacts, retry the same route, regenerate variants, accept, place, replace a version, or keep exploring.
Nothing enters a scene by accidentBring in common model formats, adjust geometry and materials, animate, validate, and return project-ready resources.
Import → edit → validate → exportTrack logical identity, versions, dependencies, reusable definitions, variants, reimports, impact previews, and rollback.
Durable project truthWe’re developing graphics workspaces for creating and refining assets alongside the game studio. Explore the visual direction, then bring supported results into your project.
Vector-oriented tools for paths, shapes, and composition. Build the visual language of a game without making every asset a generated result.
Sprite and image workflows for exploring an asset’s appearance. Local tools and generation research complement manual creation; availability varies by surface.
ForgeWeld brings model import, geometry, materials, animation, and output into the creation story. Additional 3D workflows are being developed and qualified separately.
Illustrations show creative directions, not application screenshots. Standalone availability, supported formats, and project integration vary by tool. This is not a claim that the entire graphics suite is released.
We’re building toward a complete free/local product, with full-quality export and access to your owned work independent of a subscription. Optional cloud services should add capacity and convenience.
Use local tools and supported models on your hardware. Local generation depends on installed models and available memory; individual providers and export targets have their own requirements.
Choose compatible hosted providers or bring your own API keys when useful. Third-party API charges are separate from any future PyreForge hosted credits.
This is the product direction, not a claim that every local workflow or desktop release is available today. Current access remains a private Chromium web beta.
PyreForge is built on Anvil, our provider-neutral intelligence and execution architecture. More than a copilot: shared boundaries connect deterministic tools, supported local models, bring-your-own API providers, and compatible hosted intelligence to the same creative project.
Choose compatible local or hosted intelligence for the work at hand, subject to available models, configured providers, and your permissions.
Tools, models, and providers sit behind shared execution contracts. A generation provider does not become the owner of your project state.
The architecture connects intelligent workflows with relevant scenes, assets, source, plans, and history—not just an isolated chat.
Execution records and artifact provenance support inspection of requested work, resolved execution, and generated results before deliberate acceptance.
These are architectural foundations, not a claim that every workflow is shipped. Availability varies by beta surface and provider. Adaptive routing and cost-optimized model selection remain research, not production capabilities.
Website scope reviewed October 6, 2026. The beta claims below retain their September 9 review boundary; newer focused development checks do not recertify the complete release.
Durable plans, reviewed Markdown import, and evidence-aware progress have passed development gates. Live conversational planning context and the complete beginner journey remain in development.
Focused acceptance is not full release certificationAuthor placeholders and preview replacing their appearance with reviewed assets. Supported replacements preserve entity identity, transforms, and gameplay bindings, with collision changes reviewed separately.
Build gameplay before final artMulti-scene editing, active-scene Play, command history, save slots, cold reload, recovery, and portable export.
Cross-workspace browser journeysDurable sprite and terrain candidates, exact receipts, acceptance, placement, reimport, versions, variants, and rollback.
More creation routes are converging nowIntegrated code editing, sandboxed execution, terminal workflows, visual graphs, and project-aware context.
Advanced tools without a separate projectLocal, BYOK, and managed routes depend on installed models, configured providers, connectivity, and each provider's capabilities.
Preflight and route receipts stay visibleTerrain painting, procedural maps, limited managed real-world DEM capture, model import, editing, animation, and output.
Coverage and scale vary by workflowProject, simulation, asset, model, and map outputs are available. Godot, Unity, desktop, and mobile are not universal release claims.
Verify the exact target before shippingThe current verified release boundary is Chromium-based web use. Other browsers and packaged desktop require separate certification.
Platform claims remain evidence-basedCreator discovery, sharing, assets, and templates belong to an optional platform track. Community and distribution features are separate from the current editor release boundary.
Independent product boundaryGame profiles provide useful starting structure and tools. They are starting points—not a promise that every genre arrives production-complete.
The first genre journeys are being proven end to end—from starter assets through creation, editing, Play, save/reload, and export. Once those full-cycle slices are polished, the same treatment will expand across tower defense, roguelike, survival, horror, racing, city builder, dating sim, educational, and the rest of the profile catalog.
The immediate priority is a complete game creation journey. Beyond that, we’re exploring ways for creators to learn, share, and shape what comes next.
SQUISH is a proposed showcase game and Workshop: a way to explore community content made through the same public tools as first-party content.
Curated game seasons and jams could turn shared ideas into small, coherent games. Events, dates, sponsorships, and prizes have not been announced.
Design, fabrication, and broadcast workflows are longer-term directions. They have separate readiness gates and are outside the current game-studio release scope.
PyreForge is a founder-led startup building a game creation studio with room to grow. Our direction combines hands-on creative tools, local workflows, and optional cloud intelligence—while keeping the project at the center.
Interested in the private beta, a partnership, or the technology behind the studio? Reach the founder directly.
founder@pyreforge.devJoin our Discord for private-beta access information and help shape PyreForge. The current beta targets Chromium browsers; supported generation routes may need local models or your own provider credentials.
Join the Private Beta