Perimeter breach detection
- Person climbing fence
- Fence cut or damaged
- Hole in perimeter fence
- Object thrown over fence
AI agents
Visual AI for every drone flight.
An AI agent built for the kind of site you operate, running on your fleet, learning from your operators, and surfacing the events that matter into the alert panels, inboxes, and tools your team already lives in every day.
Agent suite
An agent for every site.
From solar farms and mines to refineries, ports, substations, and highways, each agent ships with hundreds of detection events grouped by category, and you can switch between Inspection, Security, and Incident to see what every one of them already covers.
Surveillance across the entire site.
Access point monitoring
- Gate open
- Door propped open
Armed threat detection
- Person carrying firearm
- Person carrying knife or blade
- Person carrying blunt weapon (bat, pipe, rod)
- Aggressive posture or threatening behavior
- Person wearing mask or face covering
Vandalism detection
- Person spray painting surface
- Person damaging property
- Person tampering with equipment
Theft detection
- Person carrying large item away from building
- Person loading items into vehicle
- Open container door (in yard or lot)
- Person photographing facility
Crowd anomaly detection
- Large crowd forming (10+ people)
Suspicious behavior detection
- Person pacing or circling building
- Person approaching multiple vehicles
- Person following another person
- Vehicle circling premises
- Vehicle conducting slow pass-bys
- Person looking into parked vehicles
- Person trying multiple car door handles
Site-specific intelligence
Tuned to your site. Sharper every flight.
Most computer vision runs the same way on day one as a year later. False positives stack up, operators stop trusting the agent, and the program stalls at the parts of the site that matter most.
These agents take the opposite path. Every correction your operator makes settles into the agent's baseline for your site, so the same false positive does not repeat and the agent watching your site by the ninety-day mark is meaningfully sharper than the one we deployed on day one, tuned to your assets, your access patterns, and your operational rhythm.
Always on
Configure once. It watches the rest.
From the next mission onward, the agent picks up the events that matter, captures the evidence, and lands the report in your inbox, whether the mission is a day-shift patrol, a night-shift sweep, a scheduled inspection, or an alarm response.
Custom detection
Detect anything in your own words.
Write what you want monitored and the agent picks it up, whether that is "damaged mounting structure", "crowd above 50 people", or "vehicle stopped at Gate 3 for over a minute". The new event joins the agent's set, carries across every flight from that point on, and never needs a model retrain or an engineering ticket to stay in place.
When conditions shift mid-flight, drop a one-off detection into the operations view and the agent picks it up on the next sample, then forgets it the moment the mission ends so your saved setup stays clean.
Common questions
Answers before you ask.
What can the agents actually detect?
Hundreds of pre-built events across ten agents. Perimeter breaches, armed threats, panel defects, fluid leaks, PPE violations, and more. Plus unlimited custom events you describe in your own words.
How accurate is it?
Pre-built events ship with reliability we commit to at deployment. Custom events get a 30-day baseline on your site before they go live. Accuracy compounds from there as operator corrections feed back in.
How is this different from VMS analytics?
VMS analytics watch fixed cameras. The agents watch moving aerial feeds from your drone fleet. They integrate with Genetec and Milestone, so a VMS alarm can auto-launch a verification mission and your security team works from one interface.
Where is the data processed and stored?
Captured media and detection records sit in your FlytBase gallery, processed in your account's region. Your data never trains the underlying models. Your corrections feed only the agent running on your site.
Start with the outcome
Agree on the KPIs. We'll measure them at 90 days.
Tell us the outcomes that count for your program, whether that is fewer missed events, faster response on the ones that land, or cleaner reports on the desk before the shift ends, and we'll work back from there together.