Purpose
DINO-X-MCP is the official Model Context Protocol (MCP) server provided by DINO-X. Built on the DINO-X and Grounding DINO models, it brings real-world visual perception capabilities to LLMs and multimodal applications, including APIs for image object detection, localization, and description.
Key Capabilities
- Fine-grained understanding: supports full-image detection, object detection, and region-level description.
- Structured output: returns object categories, counts, locations, and attributes, suitable for VQA (Visual Question Answering) and multi-step reasoning tasks.
- Composability: can seamlessly collaborate with other MCP servers to build end-to-end visual agents or automated pipelines.
Usage
Transport Modes
DINO-X MCP supports two transport modes: STDIO and Streamable HTTP.
Prepare an MCP Client
Any MCP-compatible client can be used.
Obtain an API Key
Apply for an API Key on the DINO-X platform (new users can get a free quota).
Configure MCP
Add the configuration to your MCP client configuration and replace it with your own API Key. First install Node.js: you can download the installer from nodejs.org, or install it using a command. Then configure the MCP client. Note: please replace your-api-key-here with the real key. Execute after confirming that Node.js is installed.
CLI Flags and Environment Variables
Common flags:
- --http: start in Streamable HTTP mode (otherwise defaults to STDIO)
- --stdio: force STDIO mode
- --dinox-api-key=...: set the API Key
- --enable-client-key: allow passing the API Key via ?key= in the URL (Streamable HTTP mode only)
- --port=8080: HTTP port (default 3020)
Environment variables:
- DINOX_API_KEY (required/conditionally required): DINO-X platform API Key
- IMAGE_STORAGE_DIRECTORY (optional, STDIO): directory for saving annotated images
- AUTH_TOKEN (optional, HTTP): if set, the client must send Authorization: Bearer <token>
Configuration examples: client configuration when using ?key=; how to use AUTH_TOKEN with a gateway that injects Authorization: Bearer <token>; and a client example using supergateway.
Supported Image Sources
- STDIO: file:// and https://
- Streamable HTTP: only https:// is supported
Supported Image Formats
jpg, jpeg, webp, png.
Development and Debugging
- During development, watch mode can be used for automatic rebuilds.
- MCP Inspector can be used for debugging.
Applicable Scenarios
Suitable for scenarios that need to add visual perception capabilities to multimodal applications or AI agents, such as object detection, image understanding, visual question answering, multi-step reasoning, and collaborating with other MCP servers to build visual agents or automated pipelines.
Limitations
The above information does not provide additional limitation notes. The license is Apache License 2.0.
