Platform Platform OverviewArchitecture
Jurisdictions Thermal InspectionPrecision MappingFire MonitoringISR Operations
Papers Partner with us
Active Development · Open Source · SBIR Phase I Targeting 2026

Mission
Governance
Platform

AI can propose. FlightLaw decides. The record proves it.

Flightworks is the governance and evidence layer for autonomous drone operations.

Deterministic rules, edge enforcement, replayable audit trails, and principal authorization for mission-critical autonomy. Built for hardware partners, evaluators, and defense/public-safety teams who need to prove what happened, why it happened, and that it will happen the same way again.

4+
Governed Use Cases
100%
Edge · No Cloud
0
Black-Box Decisions (Design Goal)
Sentinel Core · Live Authority Gate
MAVLink frame evaluation · BVLOS corridor · hash-chain audit
Intercepting
Last verdict
AUTO_ALLOW
Within authorized bounds
Frames evaluated
3,041
Session total · 100% hash-chained
Tamper-evident audit chain
What exists today
Development state
is explicit here.
Complete
FlightLaw architecture
SwiftVector / RustVector spec
Sentinel relay (Rust)
ThermalLaw PRD
SurveyLaw PRD
FireLaw PRD
ISRLaw PRD
In Progress
Watch Station native app
MAVLink direct connection
Core ML inspection models
Field workflow prototype
Next
Alpha evaluator program
SBIR Phase I submission
Hardware partner field demo
Pendleton training eval
Open source

Flightworks Control is built on two open source foundations: the SwiftVector enforcement kernel and the GeoVector geospatial layer. Both are available under MIT license. Evaluation starts with the code.

GitHub →
Who this is for
SBIR & Defense Evaluators
DoD · Public Safety · Program Offices

You need governed autonomy with pre-loaded authority, degraded-mode discipline, and tamper-evident audit trails. ISRLaw and FireLaw were designed for exactly this problem.

Hardware & Platform Partners
OEMs · Integrators · Platform Operators

You have drone hardware or a flight platform and need a governance layer that doesn't compete with your GCS. Flightworks governs the mission. You fly it. MAVLink-compatible from day one.

Technical Evaluators
Architects · Engineers · Researchers

The SwiftVector kernel and GeoVector spatial layer are open source under MIT. The architecture papers are public. Start with the SwiftVector spec and the Conformance Contract. Evaluation starts with the code, not the pitch.

Field Validation Partners
Test Ranges · Training Grounds · Operational Teams

You have access to approved airspace, MAVLink-compatible hardware, or operational contexts where governed autonomy needs field-level validation. We bring the governance kernel. You bring the aircraft and the mission.

Start the conversation →
Why it matters

Autonomous ops without
governance aren't defensible.

What Flightworks delivers

Defensible post-mission review. Every finding has a chain of custody. Every principal decision is timestamped and hash-chained. If a regulator, insurer, or commander asks what happened — you can prove it, replay it, and hand them the evidence packet.

Safer autonomy boundaries. AI operates within the authority envelope you pre-authorize. Laws enforce the boundaries architecturally — not as bolted-on checks, but as constitutional constraints that cannot be bypassed.

Cleaner separation between AI and authority. The AI proposes. You authorize. The audit trail records both. Accountability is never ambiguous.

Evaluator-ready artifacts. Open-source kernel, deterministic replay, SHA256 audit chain, and field-tested workflows. Technical evaluation starts with the code.

The Opportunity

Build operator confidence without slowing the mission.

The market gap is not autonomy — it is trust in autonomy. Operators need to know what the system decided, why it decided it, and that they can prove it after the fact. The product that closes this gap wins the regulatory race and the operator's confidence.

< 25 min
Prototype Goal: Landing to Report
100%
Design Goal: Decision Coverage
0
Design Goal: Unaudited Decisions
Architecture

One kernel.
Nine laws.
Zero black boxes.

FlightLaw is the constitutional safety kernel — nine laws (0–8) inherited by every jurisdiction. Deterministic reducers, SHA256 audit chains, and mandatory principal authorization at every risk threshold. The laws are structural — they cannot be bypassed at runtime.

FlightLaw Constitutional Stack · SwiftVector v2.0
PLATFORM MAVLink 10Hz TELEMETRY LAW 3 OBSERVATION LAW 4 RESOURCE LAW 7 SPATIAL REDUCER PURE FUNCTION DETERMINISTIC OUTPUT NEW STATE SHA256 HASH AUDIT TRAIL APPEND-ONLY HASH CHAIN TAMPER-EVIDENT DETERMINISTIC REPLAY LAW 8 · AUTHORITY PRINCIPAL APPROVAL FOR MEDIUM + HIGH RISK HARDWARE LAW EVALUATION STATE TRANSITION VERIFICATION EVIDENCE
Sentinel Edge Relay · RustVector
<5ms
Target Relay Latency

Transparent MAVLink v2 forwarding. Zero-copy parsing.

8
Deterministic Rules

Geofence, deviation, battery, altitude, time window, no-fly, RTH reserve, operator auth.

3
Authority Verdicts

AUTO_ALLOW, AUTO_DENY, ESCALATE_TO_OPERATOR.

100%
Design Goal: Replay Fidelity

Deterministic replay of any audit log. Identical results, every time.

Deep dive: Architecture & FlightLaw → Explore Sentinel →
Flightworks Control

Platform components.
Mission modules.

Two platform components provide the operational infrastructure. Mission modules are composable FlightLaw jurisdictions that load into Flightworks Control based on operational context. The same safety kernel — battery management, geofencing, operator authority, audit trail — operates identically across every mission type. The module defines the domain-specific constraints.

Platform Components
Watch Station
Principal Interface · Part of Flightworks Control

The place where the operator exercises authority over the governed AI system. Pre-flight authorization, in-mission monitoring, and post-flight evidence review. Mac, iPad, and iPhone.

Sentinel
MAVLink Edge Relay

RustVector enforcement kernel at wire speed. Evaluates every MAVLink frame against deterministic rules before it reaches the autopilot. Governance at the edge with zero cloud dependency.

Mission Modules
ThermalLaw Spec Complete
Thermal
Inspection
Post-hail · Infrastructure · Roof Assessment

AI proposes damage candidates from onboard RGB + thermal imagery. Operator approves each finding. Documentation pack generated with full chain of custody. No cloud dependency, no hallucinated findings.

EvidenceLaw — valid finding structure enforcement
FindingApprovalLaw — principal gate per candidate
ReportLaw — output blocked until queue clears
SurveyLaw Spec Complete
Precision
Mapping
RTK · Photogrammetry · Engineering Grade

Engineering-grade spatial accuracy with deterministic verification. RTK precision enforced at 2cm horizontal. Grid adherence, GSD compliance, and gap detection run as laws — not suggestions.

GridLaw — deterministic geometric constraint
GSDLaw — altitude and capture compliance
RTKLaw — 2cm horizontal accuracy gate
FireLaw Architecture Draft
Wildfire
Monitoring
Overnight · Multi-Asset · Escalation-Tiered

First jurisdiction to exercise multi-asset governance and degraded-mode discipline. When comms fail, autonomy contracts — never expands. Hotspot detection through a 4-tier escalation model from Routine to Wake-the-IC.

EscalationLaw — 4-tier deterministic authority
TaskLeaseLaw — bounded fleet coordination
DegradedModeLaw — authority contracts on comms loss
ISRLaw Architecture Draft
ISR Swarm
Operations
Defense · Contested · Comms-Denied

Designed for environments where communication denial is the expected operating condition, not a failure state. Pre-loaded authority model. EMCON governance. Classification-aware state transitions. Every autonomous decision pre-authorized at T-minus.

EMCONLaw — RF emissions as boundary violations
ClassLaw — classification at state transition level
PreloadLaw — full authority envelope set before launch
Mission Flow

Before, during,
and after the flight.

Flightworks governs the authority envelope before launch, enforces bounded rules at the edge during execution, and governs evidence review after flight. The GCS executes the flight plan. Flightworks governs the mission.

Pre-Flight
01
Brief + Authorization
Watch Station
  • Asset layer from GeoPackage
  • WeatherKit Law 2 evaluation
  • Principal authorizes flight plan
  • Law 7 geofence corridor locked
  • Mission Package generated
In Flight
02
Execution + Edge Enforcement
GCS · DJI Pilot · MAVLink Direct
  • GCS or MAVLink direct execution
  • Sentinel enforces bounded rules at the edge
  • Telemetry logged at 10Hz+
  • FlightLaw kernel evaluating in real time
  • Every frame hash-chained to audit trail
After Action
03
Ingest · Review · Report
Watch Station
  • Images georeferenced locally
  • Core ML triage on Apple Silicon
  • Findings queue for principal review
  • Each finding: explicit approval
  • Report gated until queue clears
The handoff model

The Mission Package is a GeoPackage file — asset layer, authorized plan, mission ID. On MAVLink hardware, Watch Station executes the mission directly. No GCS required.

What makes this defensible

Every principal decision is timestamped, immutable, and hash-chained. Prototype goal: signed post-flight report in under 25 minutes. The audit trail is the product.

The Regulatory Thesis

The governance gap
is real and growing.

Proposed BVLOS rules and growing defense procurement requirements point toward the same need: bounded autonomy, traceable decisions, and reviewable evidence. Flightworks is built to meet that need architecturally.

BVLOS Rulemaking
Emerging Requirements

Proposed BVLOS frameworks point toward requirements for deterministic behavior, bounded authority, and auditable decision chains. FlightLaw is designed to satisfy these requirements architecturally — not as a compliance checklist.

Defense Procurement
Accountability Standards

Growing defense procurement requirements emphasize NDAA-compliant UAS with auditable governance. ISRLaw and FireLaw target this class of requirement — pre-loaded authority envelopes with hash-chained evidence.

Market Gap
No Governance Layer Exists

GCS platforms execute flight plans. Nobody governs them. The accountability gap between autonomy and oversight is where safety incidents, regulatory exposure, and trust failures accumulate.

The demo
is the pitch.

Technical evaluation starts with the code, the architecture papers, and the running Sentinel relay — not a slide deck. We are actively seeking hardware partners and defense-adjacent organizations to co-develop field evidence before SBIR Phase I submission.

If you have MAVLink-compatible hardware, access to approved training grounds, or a defense use case that needs governed autonomy with a traceable audit trail — that is the conversation we want to have.

Current engagement model: technical evaluations, field-validation partnerships, and SBIR-aligned co-development.

Start the conversation Read the Papers
SBIR Phase I · 2026
Defense & Public Safety

ISRLaw and FireLaw target DoD and public safety problem sets. Pre-loaded authority envelopes, degraded-mode governance, and tamper-evident audit chains address the core accountability gap in autonomous operations.

Hardware Partnership
Bring Your Platform

MAVLink-compatible drone hardware + Flightworks governance = field-demonstrable governed autonomy. We bring the kernel, the laws, and the evidence layer. You bring the aircraft and the operational context.

Technical Evaluation
Start with the Code

The SwiftVector kernel and GeoVector geospatial layer are open source under MIT license on GitHub. The architecture papers are at agentincommand.ai. FlightLaw and the Flightworks Control application suite are proprietary. No NDA required to evaluate the open source foundations.