Dragoneye

Dragoneye turns video and images into structured data using zero shot vision models, with no data labeling and no model training. A user describes what should be detected, such as a sponsor logo, a vehicle type or a piece of road furniture, and Dragoneye builds a working model from that description in minutes. Running the model on footage returns records of objects, their attributes and their activity over time, for example how long a brand logo stays on screen, how clearly it appears and what blocks it. It is aimed at developers and operations teams that need measurements from video: sponsorship exposure reports, turnaround tracking, traffic surveys and road inventories are the use cases the vendor shows. Models are defined as a schema, either by a coding agent such as Claude Code or Codex through the Dragoneye MCP server or by the agent in the browser Playground. Models can also start from ready templates, and they run through Python and Node SDKs in a few lines of code. Pricing is published and usage based. The Pay as you go tier includes $10 of free credits on sign up, then charges $0.046 per minute of video sampled at 5 frames per second and $0.005 per image. Video below 5 frames per second is billed at the 5 frame rate, and higher rates scale the charge proportionally, so 10 frames per second costs twice the base rate. Credits never expire but are not refundable. The Scaled tier is quote based and adds volume discounts, bespoke model finetuning in beta and priority support.