Send a ghost.
One agent, or the whole wing.
GhostOS runs the machine.
- ·LLM agent crew
- ·ROS 2 navigation and control
- ·Perception models
- ·Visual positioning
- ·Local world model
- ·Peer coordination
Everything a mission needs to keep going.
- ·Larger planning models
- ·Mission builder and simulation
- ·Fleet world model
- ·Operator interface
- ·TAK bridge
Where people plan, watch, and retask.
- ·View and control from anywhere
- ·Larger models over API
- ·Heavier CV and analysis
- ·Imagery, map, and data feeds
- ·Fleet view across deployments
More reach and more compute. Never a dependency.
Only the first column has to be there. Each step to the right adds reach and compute, but none of it decides for the machine.
GhostOS ships onboard GhostWing machines like the Phantom. It isn't sold as a standalone strap-on module. Want it on your own platform? The SDK is available to select integrators and OEMs, case by case, not as an off-the-shelf SKU.
An agentic crew on every craft.
Each agent reasons over a shared mission graph. They hand off cleanly and degrade gracefully when one fails.
A library of plays. Not a mass of drones.
Plays compose into missions. Vespera selects them; GhostOS executes them on the craft. The wing switches plays as the situation changes, without operator intervention. A containment that slips becomes search → reacquire → contain again.
Scale autonomy. Mission by mission.
An operator flies it directly by RC, FPV, or hand controller.
Training, demonstrations, and tightly controlled airspace.
Fixed routes and waypoints, set before launch.
Known terrain, working comms, routine patrols.
Onboard behaviors adjust in flight, including formations, follow, and stay on target.
Force protection, tracking, and coverage patterns.
The machine plans, executes, adapts, and reports on its own.
Comms denied, high tempo, long duration.
Set per mission in Vespera; GhostOS runs the level you picked onboard each machine.