vision-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@vision-mcp 看这张截图,里面有什么错误信息?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
vision-mcp
A plain-text model that can see images. Just copy and paste a screenshot—no need to save files or fill in gateways.
Default vision: Groq free qwen/qwen3.6-27b.
Installation
cd D:\workplace\vision
python -m venv .venv
.venv\Scripts\pip install -e .
.venv\Scripts\python -m vision_mcp setup --key gsk_你的keyFree key: https://console.groq.com/keys
setup will:
Write the key to
%USERPROFILE%\.vision-mcp\config.json(not into mcp.json)Automatically connect to existing Command Code / Grok / OpenCode / Cursor / Claude on this machine
Install the
vision-seeskill
Then restart the corresponding agent, or use /mcp to reconnect in Command Code.
Related MCP server: clipboard-vision-mcp
Toggle
.venv\Scripts\python -m vision_mcp disable
.venv\Scripts\python -m vision_mcp enable
.venv\Scripts\python -m vision_mcp statusUse disable when not needed. In Command Code, you can also disconnect via /mcp in the session without deleting the configuration.
Usage
Copy a screenshot (
Win+Shift+S)Tell the agent: look at this image / what's the error in the clipboard
The agent calls
vision_see, which reads the clipboard by default
[Image 1] is treated as clipboard content, not a file path.
Tools
Tool | Description |
| View an image. Defaults to clipboard, also accepts data URI / base64 / URL / path |
| Check configuration and whether the clipboard contains an image |
Change Model
Only modify %USERPROFILE%\.vision-mcp\config.json or set an environment variable. No need to touch the MCP configuration.
{
"api_key": "gsk_...",
"model": "qwen-vl-max",
"base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1"
}Manual Configuration
If setup has already been run, there's no need to paste it again. The format is the same as Command Code:
{
"mcpServers": {
"vision": {
"transport": "stdio",
"enabled": true,
"command": "D:\\workplace\\vision\\.venv\\Scripts\\python.exe",
"args": ["-m", "vision_mcp"]
}
}
}Do not write the key into this file.
Available Tools
2 toolsvision_seeARead-onlyIdempotent
Look at an image and return a text description/answer. You cannot see images yourself — you MUST call this. Default source is the OS clipboard (user copied or pasted a screenshot). Do NOT ask the user to save a file. Call immediately when the user pastes an image, mentions screenshot/clipboard/图片/截图, or you see an [Image] placeholder. image: omit/'clipboard'/data URI/raw base64/https URL/local path (last resort). question: what to extract or answer.
Inputs: image, question, max_tokens. Returns: result
vision_statusARead-onlyIdempotent
Check vision config and whether the OS clipboard currently holds an image. Use when vision_see fails or before asking the user to copy again.
Inputs: no named parameters. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 2 tools
The two tools have completely separate purposes: 'vision_see' performs the core image analysis and description, while 'vision_status' checks configuration and clipboard state. There is no overlap or ambiguity in their roles.
Both tools follow a consistent 'vision_verb' naming pattern: 'vision_see' and 'vision_status'. The prefix establishes the domain clearly, and the verbs are distinct and descriptive.
With only 2 tools, the server is on the borderline of being too thin for a typical MCP server. While the two tools cover the essential workflow, the set feels minimal and lacks ancillary tools that might be expected (e.g., configuration or image management).
The server covers the core use case of analyzing an image and checking readiness. Minor gaps exist (e.g., no explicit error recovery or format listing), but agents can work around these by combining the existing tools or relying on user assistance.
Maintenance
Related MCP Connectors
Shared memory for AI coding agents. Save once, reuse from Cursor, Claude Code, Codex.
Connect ChatGPT, Claude, Cursor or VS Code to your Windows PC: files, terminal, browser, apps.
Private memory for AI agents: save and recall facts across Claude Code, Codex, Cursor and ChatGPT.
Turn calls and notes into process flow diagrams in Claude, ChatGPT, VS Code, Cursor, and Teams.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceProvides free and unlimited vision capabilities for AI coding assistants using the Groq API and Meta Llama 4 Vision model. It enables users to perform image analysis, OCR, UI layout description, and error diagnosis directly from screenshots and documents.-
- AlicenseBqualityDmaintenanceEnables text-only AI models to understand clipboard images by describing them through a vision model, eliminating manual file saving.1242MIT
- AlicenseAqualityDmaintenanceGives vision-less LLMs the ability to recognize clipboard screenshots and images by proxying to an OpenAI-compatible vision model.131 npm2MIT
- AlicenseAqualityCmaintenanceGives text-only LLM coding agents vision by routing images to a multimodal model and returning detailed textual descriptions. Supports local files, URLs, clipboard, base64, raw bytes, and multiple providers like OpenAI, Anthropic, and Gemini.174 npm13MIT