Plug into MCP

This page shows how to configure HiAPI MCP. Add a HiAPI service in your AI tool’s MCP / Connectors settings, enter the remote URL and your own HiAPI API key, then generate images, edit images, create videos, or inspect models directly from chat.

Quick Setup

1. Get your API key

   Go to HiAPI Dashboard → API Keys and create or copy an API key.

2. Add a new MCP service in your AI tool

   Open the MCP, Connectors, Tools, Integrations, or Servers settings in your AI tool. Add a new service and name it hiapi.

   Fill in these fields:

   Field	Value
   Name / Server name	hiapi
   Type / Transport	Remote, HTTP, or Streamable HTTP
   URL	https://mcp.hiapi.ai/mcp
   Header name	Authorization
   Header value	Bearer YOUR_HIAPI_API_KEY

3. Enter the remote URL

   https://mcp.hiapi.ai/mcp

4. Add the authorization header

   Authorization: Bearer YOUR_HIAPI_API_KEY

5. Save and restart the AI tool

   Most AI tools load MCP settings at startup. Fully quit the tool, then open it again.

6. Verify the connection

   Ask:

   Use HiAPI to list available image models, then generate a product poster with gpt-image-2.

JSON Config Template

If your AI tool accepts JSON configuration, use this template:

{
  "mcpServers": {
    "hiapi": {
      "url": "https://mcp.hiapi.ai/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_HIAPI_API_KEY"
      }
    }
  }
}

If the client provides a UI, use the same fields:

Field	Value
Name / Server name	hiapi
Type / Transport	Remote, HTTP, or Streamable HTTP
URL	https://mcp.hiapi.ai/mcp
Header name	Authorization
Header value	Bearer YOUR_HIAPI_API_KEY

Let an AI Agent Configure It

If you want an AI tool to write the config for you, give it this page or llms.txt and say:

Use the HiAPI MCP docs to add the remote MCP URL and Authorization header to the current tool’s MCP config. After restart, verify the connection with list_models and generate_image.

How to Ask the AI

After connection, you do not manually fill tool fields. Describe what you want in chat, and the AI tool will choose the HiAPI MCP tool and fill the right parameters for the selected model.

Goal	Example request
Generate an image	Use gpt-image-2 to generate a 16:9 product poster with a clean layout and Chinese headline
Generate with reference images (selected models only)	Use Nano-Banana-2 with my uploaded image to generate a Japanese magazine cover
Edit an image (requires an image-editing model)	Use Nano-Banana-Pro to edit https://example.com/source.png and replace the background with snowy mountains
Generate a video	Use seedance-1-5-pro to create a 5-second 1920*1080 coffee ad video
Animate an uploaded local image (requires image-to-video)	Use wan2.7-i2v to animate my uploaded image into a 5-second landscape video
Browse models	List the currently available HiAPI image and video models
Check model parameters	Check which aspect ratios and resolutions Nano-Banana-Pro supports
Check pricing	Check pricing and multipliers for gpt-image-2

If you name a model, the AI tool should use your choice. If you do not name one, ask it to inspect the model list and capabilities before recommending a model.

Important: uploaded-image features only work with selected models

• Nano-Banana, Nano-Banana-2, and Nano-Banana-Pro support optional reference images. Without uploaded images they work as text-to-image models; with uploaded images they can generate from references, restyle, or edit.
• Nano Banana reference images support up to 14 images in JPG, PNG, or WebP. Keep each image under 10MB. Uploaded images are compressed to a maximum side length of 2048px before the API call.
• edit_image is only for Nano Banana models and requires at least one source image. For reference-image generation, the AI can also call generate_image with image inputs.
• Animating an uploaded local image is only for image-to-video models. Currently, wan2.7-i2v requires an image and seedance-2-0 accepts one optionally. First/last frames support JPG, PNG, or WebP, max 10MB each.
• Text-to-image models such as gpt-image-2, gpt-image-2-beta, qwen-image-2.0, and flux-1.1-pro do not support uploaded reference images and should not be used with edit_image.
• Remote MCP cannot read local file paths on your computer. If you upload an image in an AI tool, the AI tool should pass the image as data:image/...;base64,... or base64 content, not as a local path such as /Users/.../image.png or C:\....

Advanced: field name reference (optional)

The AI tool fills these fields automatically. You only need this section when debugging an MCP call or asking an AI agent to inspect its configuration. For normal use, describe your request using the examples above.

• prompt: generation, editing, or video creation instructions.
• model: set when the user names a model, such as gpt-image-2 or Nano-Banana-Pro.
• image_url, image_data_uri, image_base64 / image_mime_type, images: use only for reference images, image editing, or image-to-video.
• aspect_ratio, image_size, size: control image ratio, resolution, or video format. Available values depend on model capabilities.
• seconds: video duration, only for video models.
• wait_for_completion: whether video generation should wait for the completed result.

If the selected model does not support a requested parameter, the AI should check model capabilities first, then use a supported ratio, size, quality tier, or workflow.

Install for Claude Desktop

1. Get your API key

   Go to HiAPI Dashboard → API Keys and create or copy an API key.

2. Open remote connector / MCP settings

   In Claude Desktop, look for Connectors, MCP, Integrations, or Tools settings. The exact entry point may differ by version.

3. Add HiAPI Remote MCP

   • URL: https://mcp.hiapi.ai/mcp
   • Header: Authorization
   • Value: Bearer YOUR_HIAPI_API_KEY

4. Restart Claude Desktop

   Fully quit Claude Desktop, then reopen it.

5. Try it

   Type in Claude:

   Use HiAPI MCP to list available image models, then generate a cinematic mountain lake photo at sunset with gpt-image-2.

Install for Claude Code (CLI)

Use the Remote MCP configuration entry in Claude Code and point it at the HiAPI endpoint.

1. Add the remote URL

   In the Claude Code MCP configuration entry, use:

   {
     "mcpServers": {
       "hiapi": {
         "url": "https://mcp.hiapi.ai/mcp",
         "headers": {
           "Authorization": "Bearer YOUR_HIAPI_API_KEY"
         }
       }
     }
   }

2. Restart Claude Code

   Fully exit the current Claude Code session, then start it again.

3. Verify

   Use HiAPI MCP to list models, then generate a cyberpunk city image with flux-1.1-pro.

Install for Codex (CLI / Desktop)

Use the Remote MCP configuration entry in Codex CLI or Codex Desktop and point it at the HiAPI endpoint.

1. Add remote MCP config

   In the Codex MCP configuration file or UI, use:

   {
     "mcpServers": {
       "hiapi": {
         "url": "https://mcp.hiapi.ai/mcp",
         "headers": {
           "Authorization": "Bearer YOUR_HIAPI_API_KEY"
         }
       }
     }
   }

2. Restart Codex

   • Codex CLI: exit the current codex session, then start it again.
   • Codex Desktop: fully quit and reopen the desktop app.

3. Try it

   Use HiAPI MCP to list image models and generate a developer workstation illustration with Nano-Banana-Pro.

Install for Cursor

1. Open Cursor Settings.

2. Find MCP, Tools, or Integrations.

3. Add a new MCP Server.

4. Configure the remote MCP server:

   Field	Value
   Name	hiapi
   Type / Transport	Remote, HTTP, or Streamable HTTP
   URL	https://mcp.hiapi.ai/mcp
   Header	Authorization: Bearer YOUR_HIAPI_API_KEY

5. Save and confirm the connection is active.

6. Open Cursor Chat (Cmd/Ctrl + L) and say:

   Use HiAPI MCP to list image models, then generate a futuristic city image.

Install for VS Code / GitHub Copilot

Use a Remote MCP configuration in VS Code / GitHub Copilot and point it at the HiAPI endpoint.

1. Open your project in VS Code.

2. Open MCP settings or .vscode/mcp.json.

3. Use a config like:

   {
     "servers": {
       "hiapi": {
         "type": "http",
         "url": "https://mcp.hiapi.ai/mcp",
         "headers": {
           "Authorization": "Bearer YOUR_HIAPI_API_KEY"
         }
       }
     }
   }

4. Open the Command Palette (Cmd/Ctrl + Shift + P) and run the relevant MCP List Servers / Start Server command.

5. Switch Copilot Chat to Agent Mode.

6. Ask:

   List HiAPI image models and generate a developer-tool style illustration.

Install for Cherry Studio

Add HiAPI as a Remote MCP server in Cherry Studio.

1. Open Cherry Studio settings.

2. Find MCP Server.

3. Click Add server.

4. Use:

   Field	Value
   Name	hiapi
   Type	Remote, HTTP, or Streamable HTTP
   URL	https://mcp.hiapi.ai/mcp
   Headers	Authorization: Bearer YOUR_HIAPI_API_KEY

5. Save, then enable the hiapi MCP server in the chat box.

6. Ask:

   Generate a minimal tech avatar background with HiAPI.

Install for other MCP hosts

If your AI tool supports remote MCP, connect HiAPI with a remote URL and one authorization header.

1. Confirm remote MCP support

   In the target tool, look for MCP, Tools, Extensions, Servers, or Integrations and confirm it supports Remote, HTTP, or Streamable HTTP.

2. Add a new MCP server

   Use:

   Field	Value
   Name / Server name	hiapi
   Type / Transport	Remote, HTTP, or Streamable HTTP
   URL	https://mcp.hiapi.ai/mcp
   Header	Authorization: Bearer YOUR_HIAPI_API_KEY

3. Use JSON when required

   {
     "mcpServers": {
       "hiapi": {
         "url": "https://mcp.hiapi.ai/mcp",
         "headers": {
           "Authorization": "Bearer YOUR_HIAPI_API_KEY"
         }
       }
     }
   }

4. Save and restart the AI tool

   Most MCP clients load server config at startup. Fully quit the app, then open it again.

5. Verify the connection

   Ask:

   List the current HiAPI MCP models and recommend one for a Chinese poster.

   A working setup should call list_models or get_model_capabilities and return image/video model options.

Troubleshooting

Symptom	Fix
401 or unauthorized	Check that the header is Authorization: Bearer YOUR_HIAPI_API_KEY
No tools appear	Save the config and fully restart the MCP client
Model parameters are unclear	Ask the AI to call get_model_capabilities first
You want live pricing	Ask the AI to call get_pricing
Uploaded local image fails	Do not pass a local file path; ask the AI to pass image_data_uri or image_base64

---