Neura Market
IDEA-Research/DINO-X-MCP logo

IDEA-Research/DINO-X-MCP

Free

让 AI 具备细粒度的图像理解能力：识别、定位、描述你看到的任何目标。

FreeFree tier
Type
Open Source
Company
IDEA-Research
Visit Website GitHub (40 stars)

About IDEA-Research/DINO-X-MCP

DINO-X MCP is the official Model Context Protocol server from IDEA Research, powered by the DINO-X and Grounding DINO models. It empowers LLMs and AI agents with fine-grained real-world visual perception capabilities, including object detection, localization, and image captioning. The server provides structured outputs such as object categories, counts, bounding boxes, and attributes, enabling complex visual question answering and multi-step reasoning tasks. It offers two transport modes—STDIO for local or cloud runtime with file input support, and Streamable HTTP for cloud-only operation—and is designed to integrate seamlessly with other MCP servers for building end-to-end visual automation pipelines. New users receive free API quota on the DINO-X platform, and the server is open source under the Apache 2.0 license.

Key Features

Fine-grained full image detection and object localization
Region-level descriptions and structured outputs (categories, counts, bounding boxes, attributes)
Composable with other MCP servers to build end-to-end visual agents or automation pipelines
Two transport modes: STDIO (local/cloud) and Streamable HTTP (cloud)
Supports input via file:// and https:// URLs
Visualization support for annotated images in STDIO mode
New users get free API quota on the DINO-X platform
Open source and easy to set up with npm or from source

Pros & Cons

Pros
  • Provides fine-grained understanding beyond basic image classification
  • Structured outputs (categories, counts, locations) ready for downstream reasoning
  • Easy integration with the MCP ecosystem and existing LLM workflows
  • Free API quota for new users to get started quickly
  • Open source with clear setup instructions and multiple deployment options
Cons
  • Requires an API key from the DINO-X platform, creating a dependency on an external service
  • Local setup requires Node.js installation and environment configuration
  • Streamable HTTP mode does not support saving annotated images locally

Best For

Enabling LLMs to perform visual question answering with object-level understandingMulti-step reasoning tasks that combine visual detection and language processingBuilding automation pipelines for image analysis, inventory counting, or quality inspectionIntegrating fine-grained object detection into AI applications and chatbots

FAQ

How do I get an API key for DINO-X MCP?
Apply for an API key on the DINO-X platform at the provided link. New users receive free quota.
What transport modes does DINO-X MCP support?
It supports two modes: STDIO (default, for local or cloud runtime) and Streamable HTTP (cloud-only with streaming responses).
Can I run DINO-X MCP entirely locally?
Yes, you can run it locally using the STDIO transport mode with a local image directory and Node.js installed. The server can also be run from source.