DeepSeek Vision MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| vision_analyzeB | Analyze one or more images with a custom prompt. Image sources may be:
Use detail="low" for cheap/fast coarse inspection, or "original" for screenshots, OCR, charts, and other fine-detail tasks. |
| vision_describeC | Describe the visible content of an image faithfully and concisely. |
| vision_ocrA | Read visible text from a screenshot/document image. Returns the transcription as model text. Prefer detail="original". |
| vision_compareB | Compare 2+ images in the order provided. |
| vision_locateA | Best-effort semantic localization of a visible target. Coordinates are model-estimated and MUST be validated before destructive or high-impact UI actions. This is useful for agent grounding but is not a deterministic object detector. |
| vision_uploadA | Upload a local image to DeepSeek Files API and return its reusable file_id. Local paths must be inside DEEPSEEK_VISION_ALLOWED_ROOTS. expires_seconds: 3600..2592000 (1h..30d), or null for permanent storage. |
| vision_files_listA | List uploaded DeepSeek user_data image files, newest first. |
| vision_files_deleteC | Delete an uploaded DeepSeek Files API image. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 8 tools
Each tool targets a distinct visual task: custom analysis, faithful description, OCR, comparison, semantic localization, upload, listing, and deletion. The overlaps between analyze and describe are minor and clearly differentiated by purpose and usage notes.
Most tools follow a vision_<verb> pattern (analyze, describe, compare, locate, upload), but vision_files_list and vision_files_delete reverse the noun-verb order, and vision_ocr is an abbreviation noun. The shared prefix provides some consistency, but the mixed conventions prevent a higher score.
Eight tools is a well-scoped number for a vision MCP server, covering both image analysis and file lifecycle management without unnecessary bloat. Each tool serves a clear purpose.
The tool surface covers the major vision tasks (analysis, description, OCR, comparison, localization) and includes full file management via upload, list, and delete. There are no obvious dead ends; images are immutable so update/delete semantics are appropriately handled.