Skip to main content

Connect to the Deep Art AI MCP server

These examples show how to connect common MCP clients to the Deep Art AI MCP server using a Deep Art API key. The server uses the Streamable HTTP transport and exposes its MCP endpoint at https://api.deepart.ai/mcp.

For the available tools, parameters, output fields, and credit costs, see the Deep Art MCP server tools reference.

Before connecting​

You need a Deep Art API key with a positive credit balance. Every MCP request must include this HTTP header:

X-API-Key: <api-key>

Replace <api-key> with your Deep Art API key. Do not commit the key to source control. Prefer environment variables or password-protected client inputs where the client supports them.

Connect Claude Desktop​

Claude Desktop can reach the Streamable HTTP endpoint through a local stdio bridge such as supergateway. Add this entry to claude_desktop_config.json:

{
  "mcpServers": {
    "deep-art-ai": {
      "command": "npx",
      "args": [
        "-y",
        "supergateway",
        "--streamableHttp",
        "https://api.deepart.ai/mcp",
        "--header",
        "X-API-Key:<api-key>"
      ]
    }
  }
}

Replace the placeholder before restarting Claude Desktop. This configuration stores the API key in the local configuration file, so protect the file and do not commit it. supergateway is a third-party bridge and requires Node.js and npx.

Connect Claude Code​

Add the remote server with the API-key header:

claude mcp add --transport http deep-art-ai https://api.deepart.ai/mcp \
  --header "X-API-Key: <api-key>"

The command saves the API key in Claude Code's MCP configuration. Use claude mcp list to check the connection status.

Connect Codex CLI or the Codex IDE extension​

Codex CLI and the Codex IDE extension share MCP configuration on the same Codex host. Codex stores user-level configuration in ~/.codex/config.toml; trusted projects can instead use .codex/config.toml.

Set the API key in the environment:

export DEEPART_API_KEY="<api-key>"

Add this entry to ~/.codex/config.toml:

[mcp_servers.deep-art-ai]
url = "https://api.deepart.ai/mcp"
env_http_headers = { "X-API-Key" = "DEEPART_API_KEY" }

Codex reads the header value from DEEPART_API_KEY, so the secret does not need to be stored in config.toml. Restart Codex after changing its environment or configuration. Use codex mcp list or /mcp in Codex to check the connection.

Connect Gemini CLI​

Use the CLI command and replace the placeholder locally:

gemini mcp add --scope user --transport http \
  --header "X-API-Key: <api-key>" \
  deep-art-ai https://api.deepart.ai/mcp

Alternatively, add the server to ~/.gemini/settings.json:

{
  "mcpServers": {
    "deep-art-ai": {
      "httpUrl": "https://api.deepart.ai/mcp",
      "headers": {
        "X-API-Key": "<api-key>"
      }
    }
  }
}

Gemini CLI supports custom headers for remote HTTP servers. These examples store the selected key in Gemini's local MCP configuration, so protect the settings file. Use /mcp to check the connection.

Connect Cursor​

Set DEEPART_API_KEY in the environment that launches Cursor. Add the server to the user-level ~/.cursor/mcp.json file or the project's .cursor/mcp.json file:

{
  "mcpServers": {
    "deep-art-ai": {
      "type": "streamable-http",
      "url": "https://api.deepart.ai/mcp",
      "headers": {
        "X-API-Key": "${env:DEEPART_API_KEY}"
      }
    }
  }
}

Cursor resolves ${env:NAME} expressions in url and headers fields. Enable the server in Cursor and check its connection status.

Connect VS Code​

Add the server to .vscode/mcp.json for a workspace connection. Use a password-protected input so the API key is not stored directly in the file:

{
  "inputs": [
    {
      "type": "promptString",
      "id": "deepart-api-key",
      "description": "Deep Art API key",
      "password": true
    }
  ],
  "servers": {
    "deep-art-ai": {
      "type": "http",
      "url": "https://api.deepart.ai/mcp",
      "headers": {
        "X-API-Key": "${input:deepart-api-key}"
      }
    }
  }
}

VS Code prompts for the API key when starting the MCP server and masks the password input during entry. Start the server from VS Code's MCP server view to check the connection.

Connect Windsurf​

Set DEEPART_API_KEY in the environment that launches Windsurf. Add the server to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "deep-art-ai": {
      "serverUrl": "https://api.deepart.ai/mcp",
      "headers": {
        "X-API-Key": "${env:DEEPART_API_KEY}"
      }
    }
  }
}

Windsurf resolves ${env:NAME} expressions in serverUrl, url, and headers fields. Enable the server in Windsurf and check its connection status.

Connect another Streamable HTTP MCP client​

For a client that accepts an MCP server object, configure the production URL and API-key header:

{
  "name": "deep-art-ai",
  "transport": "streamable-http",
  "url": "https://api.deepart.ai/mcp",
  "headers": {
    "X-API-Key": "<api-key>"
  }
}

Property names vary between clients. Some clients use httpUrl or serverUrl instead of url, and some call the transport http instead of streamable-http.

Verify the connection​

After connecting, check that the client discovers these five tools:

  • create_image
  • edit_image
  • create_video
  • get_video_status
  • get_credit_balance

The server also exposes the image-card and video-card MCP App resources. Clients without MCP Apps support still receive the tool's text and structured output.

As a safe first call, ask the client:

What is my Deep Art AI credit balance?

The client should call get_credit_balance and return the creditsBalance value.

Troubleshooting​

401 Unauthorized​

  • Confirm that the URL is exactly https://api.deepart.ai/mcp.
  • Confirm that the API key is valid and has no extra whitespace.
  • Confirm that the header name is exactly X-API-Key.
  • Ensure the client sends the API key on every MCP HTTP request.

402 Payment Required​

The account has no credits or does not have enough credits for the requested generation. Image creation and editing require 1,000 credits. Video generation requires 3,250 credits per requested second. See the tools reference for the complete cost table.

The client expects SSE or stdio​

Deep Art AI exposes a Streamable HTTP endpoint, not a legacy SSE endpoint. Select http or streamable-http. For a stdio-only client, use a compatible local Streamable HTTP-to-stdio bridge.

Tools do not appear after editing configuration​

  • Validate the JSON or TOML syntax.
  • Restart the client so it reloads the MCP configuration.
  • Check the client's MCP status page or command for connection errors.
  • Confirm that DEEPART_API_KEY is available to the GUI application, not only to an unrelated terminal session.