@inliner/mcp-server

MCP server for Inliner.ai — gives AI coding agents live access to your image projects, credits, and generation.

Works with any Model Context Protocol compatible tool: Claude Code, OpenAI Codex CLI, GitHub Copilot, Gemini CLI, Cursor, Windsurf, and more.

Install

Claude Code

claude mcp add --transport stdio inliner -- npx -y @inliner/mcp-server --api-key=YOUR_API_KEY

# Or with environment variable
export INLINER_API_KEY=your-key
claude mcp add --transport stdio inliner -- npx -y @inliner/mcp-server

OpenAI Codex CLI

Add to ~/.codex/config.toml:

[mcp.inliner]
command = "npx"
args = ["-y", "@inliner/mcp-server"]
env = { INLINER_API_KEY = "your-key" }

Gemini CLI

Add to ~/.gemini/settings.json:

{
  "mcpServers": {
    "inliner": {
      "command": "npx",
      "args": ["-y", "@inliner/mcp-server"],
      "env": { "INLINER_API_KEY": "your-key" }
    }
  }
}

VS Code / Cursor / Windsurf

Project-specific (Recommended): Create .cursor/mcp.json (or .vscode/mcp.json) in your project root:

{
  "mcpServers": {
    "inliner": {
      "command": "npx",
      "args": ["-y", "@inliner/mcp-server"],
      "env": {
        "INLINER_API_KEY": "your-key",
        "INLINER_DEFAULT_PROJECT": "your-project-namespace"
      }
    }
  }
}

Global setup: Add to Cursor Settings > Features > MCP, or VS Code MCP settings:

{
  "mcpServers": {
    "inliner": {
      "command": "npx",
      "args": ["-y", "@inliner/mcp-server"],
      "env": { "INLINER_API_KEY": "your-key" }
    }
  }
}

Note: Using the env field is recommended over --api-key command-line arguments for better compatibility with MCP clients.

Preferred project behavior:

• If a tool call omits project, the server resolves it in this order:
  1. INLINER_DEFAULT_PROJECT (if set)
  2. account default project
  3. first available project
  4. "default" fallback
• This reduces repetitive "which project?" confirmations in day-to-day usage.

Tools

Resources

Example Interaction

Once installed, ask your AI agent naturally:

"Create a project called 'marketing' for my marketing team"

The agent will use create_project to reserve the namespace, then you can use it for generating images.

"Add a hero image to the landing page for my acme-corp project"

The agent will:

1. Call get_project_details to get your project config
2. Call generate_image_url with the right namespace and dimensions
3. Output the <img> tag with the correct URL, alt text, and loading attributes

Smart URL behavior:

• The server recommends concise slugs using POST /url/recommend
• Then generates with full prompt context using POST /content/generate and the selected slug
• This preserves rich prompt quality while producing readable/SEO-friendly URL paths
• generate_image_url responses include the selected slug plus alternatives to help agents pick cleaner URLs when needed

"Generate a happy duck image and save it to ./images/duck.png"

The agent will:

1. Call generate_image with the description, dimensions, and output path
2. Poll until the image is ready (up to 3 minutes)
3. Save the image to the specified file path
4. Return the URL and file path

"Create a hero image for my landing page" (using create_image alias)

The agent will:

1. Call create_image with just the description (defaults to 800x600 PNG)
2. Poll until ready and save to a sensible default location
3. Return the URL and file path

"Edit this local photo to remove the background and resize to 400x400"

The agent will:

1. Call edit_image with sourcePath pointing to the local file
2. Upload the file first (if no URL provided)
3. Apply the edit instruction
4. Resize to specified dimensions
5. Save the result

"How many image credits do I have left?"

The agent calls get_usage and reports your remaining credits by type.

API Key

Generate an API key from Account > API Keys in the Inliner dashboard. Only account owners can create and revoke keys.

Pass it via:

• Environment variable (recommended): INLINER_API_KEY — Use the env field in MCP configuration files
• Command-line argument: --api-key=YOUR_KEY — Works for standalone testing, but may have parsing issues with some MCP clients

Links

• Inliner.ai
• Tutorials
• Model Context Protocol