Install Vision MCP Server

Pick your client, copy the command, done.

claude mcp add -e "GROQ_API_KEY=${GROQ_API_KEY}" vision-mcp-server -- python -m vision_mcp_server.server

Required:GROQ_API_KEY

Environment Variables

Set these before running Vision MCP Server.

VariableDescriptionRequired

GROQ_API_KEYAPI key for accessing Groq servicesYes

Once configured, Vision MCP Server gives your AI agent access to:

analyze_imageGeneral image analysis

extract_textOCR text extraction

describe_uiUI component analysis

diagnose_errorError screenshot analysis

understand_diagramDiagram interpretation

analyze_chartChart and dashboard analysis

compare_imagesImage comparison

code_from_screenshotCode extraction from screenshots

After setup, try these prompts with your AI agent:

→Describe this image: screenshot.png

→Extract text from this document: scan.jpg

→What's wrong with this error screenshot: error.png

→Explain this architecture diagram: system-diagram.png

→Compare these two UI screenshots: old-ui.png vs new-ui.png

Prerequisites & system requirements

An MCP-compatible client (Claude Code, Cursor, Windsurf, Claude Desktop, or Codex)
Python 3.8+ with pip installed
GROQ_API_KEY — API key for accessing Groq services

Conare · memory for coding agents

Save the docs, env vars, and workflow around Vision MCP Server in Conare so Claude Code, Codex, and Cursor remember it next time.