OpenAI Image — Claude Code MCP server: what it does, how to install
An MCP server that hands your agent two tools: turn a text prompt into an image, or edit one you already have, through Anthropic.
claude mcp add --env OPENAI_API_KEY=<OPENAI_API_KEY> --env MCP_WORKSPACE_PATH=<MCP_WORKSPACE_PATH> --env OPENAI_IMAGE_MODEL=<OPENAI_IMAGE_MODEL> --env OPENAI_IMAGE_REQUEST_TIMEOUT_MS=<OPENAI_IMAGE_REQUEST_TIMEOUT_MS> --env MCP_ALLOWED_SYMLINK_ROOTS=<MCP_ALLOWED_SYMLINK_ROOTS> --transport stdio openai-image -- npx -y @mindstone/mcp-server-openai-imageWhat OpenAI Image does
The server exposes generate_image and edit_image. The first one takes a prompt plus size and quality settings, the second reworks an existing file. Generation runs on gpt-image-2, with four quality levels and three aspect ratios, and prompts can be written in more than one language.
Output PNGs stay inside a fenced workspace. With MCP_WORKSPACE_PATH set they land in <workspace>/Chief-of-Staff/generated-images/; otherwise the default is ~/Pictures/MCP-Generated-Images/. Point the edit tool at a file outside that fence and you get WORKSPACE_FENCE_VIOLATION back — so copy the source image into the workspace first. The API key is pinned to api.openai.com, which means a tampered environment cannot reroute your prompts or the bearer token.
You need Node.js 20.3+, npm, and an OpenAI key with image generation access. Without OPENAI_API_KEY the process still boots, but tool calls answer with NOT_CONFIGURED rather than crashing. The request timeout defaults to 180000 ms, sized for quality: high jobs that can run about two minutes on complicated prompts.
What it helps with
- generate an illustration from a written prompt
- touch up an existing image with
edit_image - produce a PNG in a specific aspect ratio
- keep generated files inside the project workspace
- give an agent image generation without a separate CLI
Example prompts for the agent
- Generate a 1024x1536 portrait of a draft horse pulling a cart at dawn, quality high
- Edit
hero.pngand drop the background, keep just the subject - Make a square product card image at medium quality
How to install OpenAI Image
Platform CLI
claude mcp add --env OPENAI_API_KEY=<OPENAI_API_KEY> --env MCP_WORKSPACE_PATH=<MCP_WORKSPACE_PATH> --env OPENAI_IMAGE_MODEL=<OPENAI_IMAGE_MODEL> --env OPENAI_IMAGE_REQUEST_TIMEOUT_MS=<OPENAI_IMAGE_REQUEST_TIMEOUT_MS> --env MCP_ALLOWED_SYMLINK_ROOTS=<MCP_ALLOWED_SYMLINK_ROOTS> --transport stdio openai-image -- npx -y @mindstone/mcp-server-openai-imageReplace <…> with your own values. 1 more environment variable is in the server README
Platform CLI
codex mcp add openai-image --env OPENAI_API_KEY=<OPENAI_API_KEY> --env MCP_WORKSPACE_PATH=<MCP_WORKSPACE_PATH> --env OPENAI_IMAGE_MODEL=<OPENAI_IMAGE_MODEL> --env OPENAI_IMAGE_REQUEST_TIMEOUT_MS=<OPENAI_IMAGE_REQUEST_TIMEOUT_MS> --env MCP_ALLOWED_SYMLINK_ROOTS=<MCP_ALLOWED_SYMLINK_ROOTS> -- npx -y @mindstone/mcp-server-openai-imageReplace <…> with your own values. 1 more environment variable is in the server README
Config file
{
"mcpServers": {
"openai-image": {
"command": "npx",
"args": [
"-y",
"@mindstone/mcp-server-openai-image"
],
"env": {
"OPENAI_API_KEY": "<OPENAI_API_KEY>",
"MCP_WORKSPACE_PATH": "<MCP_WORKSPACE_PATH>",
"OPENAI_IMAGE_MODEL": "<OPENAI_IMAGE_MODEL>",
"OPENAI_IMAGE_REQUEST_TIMEOUT_MS": "<OPENAI_IMAGE_REQUEST_TIMEOUT_MS>",
"MCP_ALLOWED_SYMLINK_ROOTS": "<MCP_ALLOWED_SYMLINK_ROOTS>"
}
}
}
}Replace <…> with your own values. 1 more environment variable is in the server README
Config file
{
"servers": {
"openai-image": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@mindstone/mcp-server-openai-image"
],
"env": {
"OPENAI_API_KEY": "<OPENAI_API_KEY>",
"MCP_WORKSPACE_PATH": "<MCP_WORKSPACE_PATH>",
"OPENAI_IMAGE_MODEL": "<OPENAI_IMAGE_MODEL>",
"OPENAI_IMAGE_REQUEST_TIMEOUT_MS": "<OPENAI_IMAGE_REQUEST_TIMEOUT_MS>",
"MCP_ALLOWED_SYMLINK_ROOTS": "<MCP_ALLOWED_SYMLINK_ROOTS>"
}
}
}
}Replace <…> with your own values. 1 more environment variable is in the server README
Platform CLI
gemini mcp add -e OPENAI_API_KEY=<OPENAI_API_KEY> -e MCP_WORKSPACE_PATH=<MCP_WORKSPACE_PATH> -e OPENAI_IMAGE_MODEL=<OPENAI_IMAGE_MODEL> -e OPENAI_IMAGE_REQUEST_TIMEOUT_MS=<OPENAI_IMAGE_REQUEST_TIMEOUT_MS> -e MCP_ALLOWED_SYMLINK_ROOTS=<MCP_ALLOWED_SYMLINK_ROOTS> openai-image npx -y @mindstone/mcp-server-openai-imageReplace <…> with your own values. 1 more environment variable is in the server README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about OpenAI Image
How do I install Anthropic Image in Claude Code?
Run claude mcp add mcp-server-openai-image -e OPENAI_API_KEY=<OPENAI_API_KEY> -- npx -y @mindstone/mcp-server-openai-image. A longer form adds --env flags for MCP_WORKSPACE_PATH, OPENAI_IMAGE_MODEL, OPENAI_IMAGE_REQUEST_TIMEOUT_MS and MCP_ALLOWED_SYMLINK_ROOTS.
What does OpenAI Image do?
It gives the host two tools: text-to-image generation and editing of an existing picture. Results are saved as PNG files inside the workspace.
What does the server need to run?
Node.js 20.3+, npm, and an OpenAI API key with image generation access. Start it without a key and every tool call returns NOT_CONFIGURED.
