Questions, answered.

Six weeks in which we build one of your training scenarios with your team, then hand over the project and its source. It is a fixed fee for six weeks, credited in full to a yearly development license if you continue. Ask us for the price. See Pilot program.

Every pilot runs under a written agreement with Cadarn Inc., which sets ownership, confidentiality and license terms.

Recording a session and replaying it later, from any camera, is in the engine today. After-action review, instructor tools and scorecards are built as first-party plugins and run in pilot builds; they are not in a public release yet. DIS is being built.

Same inputs, same simulation on Windows and Linux, checked every night. What the check covers, and what it does not, is on Determinism.

Not yet in a release. Online multiplayer and split-screen on one PC are being built now.

Exported builds run on Windows and Linux. Web export is in preview, and Mac is coming. The editor runs on Windows.

An engine for building training simulations with an AI assistant at your side. You describe the scenario, and your assistant builds it inside the engine: the scenes, the scripts, the instructor's controls. It runs the scenario, finds what is broken and fixes it, and you approve every change. It runs on your own machines.

Almost every editor action is also a documented tool your assistant can call over MCP, the open protocol AI assistants use to work with apps. After every write a critic checks the scene and its scripts before anything runs, and a built-in playtester plays the game toward its goals and reports where it got stuck. Each change lands in the editor's activity feed, marked with the AI that made it: open the diff, or revert it in one click.

Engines were designed for people clicking through an editor. An AI working in one writes files it cannot see run. Talos Engine gives the assistant the same tools the editor uses, a way to play and test what it built, and a fast loop: 0.8 s from launching the editor to playing your game, and a 174 MB install. The faster an assistant can try, check and fix, the more it gets right.

No. Everything works in the editor by hand. Connecting an AI agent is optional, and you choose which one. Each agent is its own product, with its own pricing.

Whichever agent you already use, as long as it supports MCP: Claude Code, Codex, Gemini CLI, Grok, Cursor and others. They all connect to the same Talos Engine MCP server. Claude Code also connects in one step from the launcher.

No. Talos Engine sends nothing to an AI unless you connect one. What your assistant sends is then covered by its own terms.

Yes: simulations, training and interactive scenes use the same physics, vehicles, water and terrain. See Public safety and Maritime. During the beta, use is non-commercial; a pilot runs under its own agreement.

Evaluating it is free during the beta, for non-commercial use. Teams that build training simulations start with a six-week pilot, then continue on a yearly development license, with the pilot fee credited. Pricing for enterprise teams is on request. See Licensing.

During the beta, the Beta tester terms are the license for evaluation: free, for non-commercial use; your projects stay yours; please do not share the installer itself. Read them in full: Beta tester terms. Commercial use runs under a pilot agreement or a development license with Cadarn Inc..

Not under the Beta tester terms. Commercial use starts with a pilot agreement or a development license.

Under the Beta tester terms, you do: they give us no ownership of the content, code or assets you create. For a pilot or a license, the written agreement sets it.

Windows 10 or 11 (64-bit), a graphics card with current drivers (DirectX 12 or Vulkan), and disk space for the launcher, one engine version and your projects. No compilers, SDKs or source code.

Windows and Linux: one .exe that runs on a Windows 10 or 11 PC with nothing installed, or a Linux build. Web export is in preview: a static folder any web host can serve, for browsers with WebGPU. Mac, iPhone and Android come later.

Paste one prompt into your AI assistant and it does the rest: Get started. Or follow the install guide yourself.

It is an early beta: expect bugs, and back up your projects. Known problems are on Known issues, and Report a bug in the editor's Help menu sends one straight to us.

Choose Talos Engine if you want an AI assistant doing real work in your engine: it is built around that, not added on. Scenarios replay with the same simulation on Windows and Linux, and it is small and quick (174 MB, 0.8 s to play). Choose an established engine today if you need consoles, a large asset store or a long track record: Talos Engine is early.

Real-time lighting and reflections, terrain from a recipe, oceans with weather, physics you can replay, cars and boats, TypeScript scripting, a profiler and automated playtesting. The full list is on Features.

Write to support@cadarn.ai, or report a bug from the editor's Help menu.