MC
Mission Copilot
Built on DSRP’s decide-before-execute agent architecture. Every recommendation is receipted and requires human countersign before execution.
Explainable AIHuman-in-the-loop
COSRYX AI — A DCS Labs Space Research Initiative
Mission Intelligence for Earth and Beyond
A research and pilot AI platform for mission planning, planetary simulations, explainable autonomy and verifiable evidence, including COSRYX-SPACETRACE and three Hybrid Moat features across a 692-test portfolio.
Status as of 28 Jul 2026: Software portfolio built and internally test-green. Validation class: simulation-validated and ground-demonstrated. Hybrid Moat: built · DoD sign-off pending. Independent external security audit: not completed. Not claimed: flight qualification, operational space deployment, certified PQ hardware root, agency acceptance.
Status legend: Built · Simulation-validated · Ground-demonstrated · Hardware-blocked · Audit not completed
Interactive Product SimulationNot Flight-QualifiedResearch and pilot interface · human confirmation required for high-impact actions
Earth · Moon · Marsplanetary mission environments
Relay + Autonomycommunications and onboard intelligence
Evidence Layersource-backed recommendations and confidence
Simulation Firstresearch-ready and human-in-the-loop
COSRYX MISSION OPERATIONS
SIM-LUNAR-01 · LUNAR RELAY SIMULATION
SIMULATION ACTIVE
MISSION SCENE
SIMULATION
Lunar South-Pole Relay
Rover support · contact window analysis
AGENT TRACE · mission_planner.py
RUNNING
01
from cosryx.mission import RelayPlanner02
planner = RelayPlanner(SIM-LUNAR-01
)03
window = planner.analyse_contact_window()04
priority = planner.rank_science_packets()05
00:08:41relay.windowContact window stable · 08:42 remaining
LATEST AI DECISION
Prioritise navigation packet A7 and anomaly report B17 before imaging batch.
Product capabilities
Designed as a premium mission-control intelligence layer
COSRYX AI turns space operations into an explainable, source-backed and simulation-driven workflow across mission planning, planetary operations, communications, anomaly handling and research assistance.
PL
Mission Planner
Constraint-aware planning with receipted decision trails for Earth, lunar, Martian and deep-space scenarios.
Timeline logicConstraint aware
RL
Relay Planner
Built on DSRP — delay-tolerant store-and-forward with light-cone timestamps. 100/100 simulation scenarios, 78 tests.
Latency modellingDownlink optimisation
AL
Autonomy Lab
Bounded autonomy with evidence, confidence and escalation policies. Connected to COSRYX-MCC (built).
Safety constraintsDecision audit
PI
Planetary Intelligence
Connected to COSRYX-LENS imagery provenance (built) — every derived product will carry a cryptographic receipt.
Earth · Moon · MarsDeep-space context
AN
Anomaly Intelligence
Connected to COSRYX-ORBIT debris tracking (built) — conjunction alerts with receipted proof of warning.
Incident logicRoot-cause support
DT
Mission Digital Twin
Represent mission assets, environment states, relay nodes and simulated operations inside a single visual operating model for researchers and operators.
3D scene logicOperational context
EV
Evidence Layer
Every recommendation shows sources, confidence, assumptions, approvals and revisions. Built on DCS R-Series — 518,336 cycles, 0 failures, 99.99% test uptime over a 30-day continuous soak. Receipts anchored on-chain. Independent external security audit: not completed.
TraceabilityFuture verification
NX
COSRYX-NEXUS
COSRYX-NEXUS — The interplanetary trust network. Signed custody receipts across delayed, interrupted links (relay); N-of-M witness signatures with dual-domain two-person control (mesh + coordination); deterministic byte-identical simulation. One product, three layers — an Earth→Moon→Mars custody chain that survives link outages.
Built — simulation-validated65 tests
ST
SWARM TRUST
SWARM TRUST — A cryptographic receipt for every autonomous-robot-swarm decision. Byzantine-tolerant (survives ⌊(n−1)/3⌋ faulty or corrupted robots), with radiation-fault (SEU) scrubbing and automatic quarantine of a rogue unit. Prove to an agency what the swarm decided, when, why, and under whose consensus — even during a communication blackout. Rides COSRYX-NEXUS + COSRYX-EDGE.
Built — simulation-validated38 tests
SA
SSA ACCOUNTABILITY
SSA ACCOUNTABILITY — Liability evidence for satellite and debris-removal operators. Prove you warned them; prove the maneuver was safe. Binds each conjunction alert to the exact observations it derived from, in the international CCSDS standard format, mapped to the 1972 Space Liability Convention. Rides COSRYX-ORBIT + COSRYX-SPACETRACE.
Built — ground-demonstrated (evidence-service)34 tests
MP
MFG PROVENANCE
MFG PROVENANCE — Certify that a fiber, material, or pharmaceutical batch was genuinely made in orbit — on this date, by this process. Manufacturing telemetry flows through the LENS processing-chain to a C2PA "made in orbit" certificate. Rides COSRYX-LENS.
Built — ground-demonstrated (provenance seam)33 tests
STX
COSRYX-SPACETRACE
SSA data provenance beneath ORBIT: sensor observations, conjunction alerts and manoeuvre decisions form a verifiable sensor → alert → spacecraft-action chain.
Built — ground-demonstrated29/29 tests
H1
COSRYX-HYBRID-01
Radiation quarantine and R+11 compute binding. Deterministic replay mismatch quarantines a record before a clean compute receipt can be emitted.
Built · DoD sign-off pending24/24 tests
H2
COSRYX-HYBRID-02
Post-quantum imagery provenance requiring both receipt legs over identical canonical bytes; downgrade and single-leg paths are rejected.
Built · DoD sign-off pending37/37 tests
H3
COSRYX-HYBRID-03
Air-gapped MCP connector producing local R+2 chained receipts without storing raw tool arguments or outputs.
Built · DoD sign-off pending28/28 tests
COSRYX Hybrid Moat
Three built features for constrained mission environments
One shared canonical receipt substrate across radiation-aware compute, dual-leg imagery provenance and air-gapped tool execution. 89/89 tests, 0 fail; DoD sign-off pending.
Anomaly quarantine before trust
Model, TEE, output and replay digests are bound into one record. Divergence causes quarantine.
Dual-leg imagery provenance
LENS-VERIFIED is earned only when both legs verify over the same canonical bytes.
Local air-gapped tool receipts
Telemetry-tool calls create local R+2 chains without network modules or raw payload retention.
The Post-Quantum (PQ) leg currently utilizes size-conformant ground stand-ins (pending formal NIST FIPS 204 ML-DSA / FIPS 205 SLH-DSA hardware security module certification). Not flight-qualified. Not independently audited by external third parties.
Premium 3D product experience
Mission console preview
An interactive software experience that feels like a premium mission-control product — not a static brochure. Choose an environment and inspect planetary, relay and operational data.
Mission Copilot
“What should we transmit first?”
COSRYX AI can answer questions like a mission analyst: what to downlink first, whether to delay a command, where the strongest uncertainty lies, and which evidence supports a recommendation.
Output is presented as operational guidance with confidence and evidence — not as unsupported automation.
Simulation Lab
Scenario Preview
Earth Mission MeshOrbital intelligence · climate observation · debris awareness
Simulation feed
Simulation Preview
Window Assessment
Relay window remains viable for 08:42.
Relay window remains viable for 08:42.
Stable
Anomaly Review
Thermal variance above nominal on node B17.
Thermal variance above nominal on node B17.
Investigating
Autonomy Guardrail
Human confirmation required before high-impact uplink.
Human confirmation required before high-impact uplink.
Required
Mission Confidence
Aggregated confidence based on telemetry, relay budget and terrain context.
Aggregated confidence based on telemetry, relay budget and terrain context.
Status as of 28 Jul 2026: Software portfolio built and internally test-green. Validation class: simulation-validated and ground-demonstrated. Hybrid Moat: built · DoD sign-off pending. Independent external security audit: not completed. Not claimed: flight qualification, operational space deployment, certified PQ hardware root, agency acceptance.
The Post-Quantum (PQ) leg currently utilizes size-conformant ground stand-ins (pending formal NIST FIPS 204 ML-DSA / FIPS 205 SLH-DSA hardware security module certification). Not flight-qualified. Not independently audited by external third parties.
Internal controlled soak campaigns (not operational space service). Example: 30-day internal soak · reported 0 failures under test conditions.
Status legend: Built · Simulation-validated · Ground-demonstrated · Hardware-blocked · Audit not completed
Open evidence pack index →Selected evidence is public; additional source records are available under NDA for qualified audit review.
Explainable mission intelligence
Every recommendation shows its evidence
COSRYX AI separates scientific sources, simulated mission inputs, assumptions and model interpretation so operators can understand why a recommendation was made before approving it.
RECOMMENDATION BUNDLE · SIM-LUNAR-01SIMULATION-VALIDATED
RecommendationPrioritise navigation packet A7 and anomaly report B17.
91%confidence
01
Telemetry stream 07Link quality and power reserve within safe simulation range.
SIMULATION FEED02
Relay-window calculation08:42 remaining with constrained bandwidth.
MODEL03
Terrain-navigation scorePacket A7 has the highest immediate operational value.
DATA04
Human approval policyHigh-impact uplink remains blocked until confirmation.
GUARDRAILCompleted infrastructure evidence
Five soak campaigns · all PASS
These software and infrastructure tests support receipt continuity, fault recovery, partition detection and cryptographic-agility claims. Independent external security audit: not completed.
Transparency: the 7-day, 15-day and 30-day raw records contain transient dependency errors that were recovered during the tests. They are disclosed in the evidence records and are not presented as test failures: each formal verdict was PASS, cycles_failed was 0, and no unrecovered halt was recorded.
0 failures · 6/6 faults · 8/8 checks
Read full test PDF →Raw JSON →0 failures · 14/14 faults · 99.95% test uptime · recovered transient fetch error
Read full test PDF →Raw JSON →0 failures · 29/29 faults · 99.99% test uptime · recovered transient Supabase timeout
Read full test PDF →Raw JSON →288 cycles · 10/10 faults · 23 checks / 0 divergence · in-test rotation
Read full test PDF →Raw JSON →637 cycles · 23/23 faults · 52 checks / 0 divergence · 8/8 partials refused
Read full test PDF →Raw JSON →692/692 COSRYX tests · 0 failFoundation 249 + S1–S6 220 + SPACETRACE 29 + S7–S9 105 + Hybrid Moat 89
Planetary intelligence
Beyond facts — operational relevance by world
COSRYX AI can combine science facts, environmental conditions and mission-relevance signals. The result is a much richer operational layer for Earth, the Moon, Mars and deep-space systems.
Research interpretation layer
Source-backed
Earth relevanceOrbital intelligence · disaster sensing · debris tracking
Moon relevanceRelay architecture · polar navigation · robotic support
Mars relevanceDelayed autonomy · terrain reasoning · habitat operations
Deep-space relevanceDTN logic · long-duration autonomy · scientific data triage
How COSRYX goes beyond public facts pages
Mission-first
Quick facts
Public scientific properties like day length, year length, gravity, atmosphere, temperature and distance.
Operational profile
Communication delay, environmental severity, power constraints, autonomy need and relay complexity.
COSRYX guidance layer
Mission relevance, scenario-based risk, suggested system emphasis and explainable intelligence recommendations.
Private, secure and explainable
Mission intelligence without surrendering control
COSRYX AI is designed for private workspaces, customer-controlled access, explicit approval gates and evidence-rich operations. Deployment claims remain clearly separated between current prototype capability and future enterprise options.
COSRYX CONTROL PLANE
Private Mission Workspace
Mission context, models, evidence and permissions organised inside a controlled operating boundary.
01
BYOK & Model Control
Use approved provider keys and routing policies while keeping mission workflows under organisational control.
Design target02
Data Boundaries
Separate customer context, simulation data and source evidence with clear retention and access policies.
Pilot scope03
Approval Guardrails
Require operator confirmation before high-impact actions, uplinks or irreversible workflow transitions.
Shown in demo04
Audit & Evidence
Record recommendation sources, confidence, assumptions, revisions and human decisions.
Shown in demoFuture enterprise direction
On-premise, sovereign and private-cloud deployment options require product engineering, security review and customer-specific implementation before being represented as operational.
POST-QUANTUM SECURITY
Quantum-safe trust infrastructure
COSRYX AI security infrastructure includes hybrid post-quantum signatures (Ed25519 + ML-DSA-65 + SLH-DSA), X-Wing key encapsulation (KAT-verified), and a state-hazard policy preventing stateful signature misuse in distributed systems.
The Post-Quantum (PQ) leg currently utilizes size-conformant ground stand-ins (pending formal NIST FIPS 204 ML-DSA / FIPS 205 SLH-DSA hardware security module certification). Not flight-qualified. Not independently audited by external third parties.
TESTED109 tests, 0 failX-Wing 17/17 KAT
Built for research and mission teams
One intelligence layer, multiple operational perspectives
Space Agencies
Mission studies, simulation, evidence review and decision-support research.
Universities & Labs
Planetary datasets, autonomous-system experiments and publishable simulation workflows.
Satellite Operators
Relay planning, anomaly analysis, orbital context and human-controlled recommendations.
Aerospace & Robotics
Digital-twin scenarios, rover autonomy and systems-integration research.
Government Research
Private mission studies, explainable AI and evidence-centric operational review.
Architecture
Built as a layered intelligence stack
COSRYX AI can grow from a research interface into a serious mission-intelligence platform with planners, worker agents, simulation layers, evidence traces and human review.
Mission Objectives
Goals, constraints, priorities and research intent define what the system is trying to achieve.
Simulation & Context
Planetary facts, mission environment, terrain, relay windows and resource models provide context.
AI Reasoning Layer
Copilot, planner, anomaly analysis and autonomy policies generate candidate recommendations.
Evidence & Review
Confidence, assumptions, source links and approval flows make the system explainable and controllable.
Mission Actions
Outputs become simulated actions, human-reviewed plans, datasets, reports or future operational workflows.
Pilot access
Define a mission-intelligence pilot
Share your research or operational use case. The form prepares a structured pilot request for the COSRYX team and clearly identifies which capability you want to evaluate.
01Focused scopeOne mission or research workflow
02Simulation firstNo claim of flight qualification
03Evidence includedSources, confidence and review states
04Human controlledApproval gates for high-impact actions
Preferred contact: research@cosryx.space
Transparent readiness labels
Research, concept, prototype, simulation, pilot, verified test and operational states must not be presented as equivalent.
Recommendations are reviewable
Confidence, assumptions, supporting evidence and human approval states remain visible in high-impact workflows.
Customer-controlled deployment direction
Production retention, model access, tenancy and security requirements must be agreed and implemented for each pilot.