Product

General Cognition Unit

An agentic autonomy stack for space systems

Earth-bound cognition limits space productivity.

Spacecraft sense far more than they can send home, and every decision about what to do next waits on a ground team, a contact window and the speed of light.

Only 2% of the hyperspectral images a Landsat 8–class satellite could capture in an orbit can be downlinked, due to bandwidth constraints.^[1]

~50%

of sols in Perseverance's first three science campaigns did not advance the campaign because they were spent waiting on Earth, on $150M a year of operations.^[2]^[3]

6–12 hours

How long EO operators have to wait to extract insights from imagery, because every image must be downlinked and processed on the ground before anyone knows what's in it.^[4]^[5]

What we deliver

GCU puts cognition onboard.

Your platform perceives, reasons and decides without waiting for Earth, so it can do more of the work it was built for.

Mission-tuned agent models

Our vision-language agents, distilled and tuned for your sensors, targets and power budget. Retask them in natural language, with no software or model update.

GCU module

The onboard unit that runs the agent. Fly our module, or run GCU on the compute you already fly.

CONTACT

Mission control for the agentic era. Tell your spacecraft what matters in natural language, and review every decision it makes.

Learn about CONTACT →

V&V and AIT services

Verification, validation, integration and test for learned autonomy, from mission simulation through hardware-in-the-loop, demonstrating that GCU meets your mission requirements before it flies.

Where GCU flies

Earth observation

Decide onboard what is worth downlinking, based on what your customers or science team care about today. Turn observations into natural-language alerts, and flag significant events nobody tasked for to maximize insight yeilds.

In-orbit inspection

Characterize unfamiliar objects in orbit, plan the next safe inspection pass, and report findings. For servicing, debris and space domain awareness missions.

Lunar surface operations

Select science targets, keep rovers productive through communication gaps, catch and diagnose faults before the next contact window, and team with astronauts in natural language.

Why GCU

Judgment where the data is.

Understanding, not just detection

Narrow classifiers find only what they were trained on. GCU reasons about what it sees, including things nobody planned for.

Retask in natural language

Change what matters from CONTACT, with no new model and no software upload.

Decisions at the source

Faster downlinks still need someone to know what to look for. GCU makes that call onboard, at the moment it matters.

Built to be certified

A deterministic layer, explained decisions and V&V from simulation to orbit. The safety case your assurance team needs is part of the design, not added later.

Your safety constraints, enforced by design.

Every agent action passes through a deterministic layer that enforces the constraints you define, so the guarantees are ones your team can verify, from simulation through hardware-in-the-loop.

Flight-critical control stays with your flight software. Attitude, propulsion, mobility, power and thermal remain under the host's command.
Every decision is explained. Each one is logged with its rationale for your team to review.
Operators stay in command. Configure the agent's privileges, override any decision or bypass GCU entirely.

Path to orbit

MASAQ-01

Our flight demonstration will fly GCU aboard a 3U CubeSat in low Earth orbit.

View MASAQ-01

Building a mission that needs onboard cognition?

See GCU in action, fly it on your next mission, or invest in what we're building.

Get in touch