Companion + command
An interruptible partner that notices, explains, speaks, proposes safe actions, and can take optional AUTO command.
Human-led play with useful intelligence inside the match.Independent project. EA has not endorsed and does not support this product.
The companion and Earth mission studio are the core experiences. Distribution, new theatres, and bounded autonomous agents make them easier to reach and stronger over time.
The plan remains an AI game companion, an Earth-to-Mission generator, and a web + launcher surface that makes both approachable. New mods and autonomous agents grow from that foundation; they do not replace it.
An interruptible partner that notices, explains, speaks, proposes safe actions, and can take optional AUTO command.
Human-led play with useful intelligence inside the match.Choose a real place, interpret its terrain, and let OpenRA build and validate an editable battlefield from that evidence.
A world-scale source of playable missions and skirmishes.The web studio lets people discover the project, build and share mission blueprints, download verified releases, and move into the native game.
One approachable doorway into the companion and creation tools.Reusable factions, units, art, voices, objectives, campaigns, and complete authored experiences.
Bounded headless matches with recorded evidence, graded outcomes, and lessons for stronger native strategy.
Current playable proof · 0.1.0-alpha.13A focused vertical slice proving the platform can move beyond a generated skirmish: two countries, faction-gated units, native bot priorities, and a source-dated Earth mission contract.
The current showcase is Red Sea 2026. The platform plan is deliberately release-agnostic so the next theatre can replace it without rewriting the product story.
Authored theatres, bounded agents, and a provider-neutral AI layer extend the same product without turning the homepage into documentation.
A three-mission theatre proves RTS AI can ship authored experiences beyond generated skirmishes.
See proof + boundary Native gameEach faction track has its own recognizable battlefield identity instead of simply recoloring legacy content.
See proof + boundary Local toolsAn agent can observe and command local OpenRA matches through a bounded tool surface.
See proof + boundary Local toolsAgent experiments can become durable learning rather than isolated playthroughs.
See proof + boundary AI layerThe game asks for named capabilities rather than hard-coding a model company or machine.
See proof + boundary