VLM Run’s Gateway gives developers a shared API for visual tasks, including reading text from documents and identifying objects in images. It uses an OpenAI-compatible chat interface, while an accompanying Model Context Protocol server makes the catalog available to coding agents and other MCP clients.
The catalog includes systems for finding object boundaries and estimating body poses, alongside document-reading and image-captioning models. The supported inputs and output formats vary by model.
VLM Run says the gateway can split large documents or videos into smaller jobs and reassemble their results. Developers can request structured output rather than building a separate response parser for every task.
The catalog lists both open-weight and proprietary models. It also uses different billing units for token-based models and services charged by image or audio duration. A shared endpoint therefore does not mean every model has the same capabilities, license or pricing basis.