JIUFENG API

Agent / MCP integration

Connect Jiufeng image generation to OpenAI-compatible clients, agents, and (soon) MCP.

Jiufeng Open API offers three paths for agents and automation.

1. OpenAI-compatible clients

The gateway follows the OpenAI-compatible style: point your client's base_url at Jiufeng and use your sk-jf-live- key to reuse an existing SDK.

Python (openai SDK)
from openai import OpenAI
 
client = OpenAI(
    api_key="sk-jf-live-YOUR_KEY",
    base_url="https://api.jiufeng.ai/v1",
)

Note: Jiufeng image generation is asynchronous. /v1/images/generations returns a task_id (not OpenAI's synchronous data[].url), so you still poll per Query tasks or receive a Webhook for the final result. Treat it as "submit a task", not "get an image in one call".

2. Machine-readable specs (llms.txt / OpenAPI)

Feed machine-readable specs to agents and code-gen tools so they can understand the endpoints and fields automatically:

  • llms.txt — an LLM-oriented site summary (product blurb + base URL + auth + endpoint list + links to each doc page) to give agents quick context on Jiufeng Open API.
  • OpenAPI spec — the complete endpoint and request/response schema (openapi.yaml, kept in sync as the API evolves).

Both are downloadable from the site root: https://docs.jiufeng.ai/llms.txt and https://docs.jiufeng.ai/openapi.yaml.

3. MCP Server (coming soon)

The official jiufeng-mcp will wrap image generation and task queries as Model Context Protocol tools, letting MCP-capable agents (Claude, various IDE agents) call Jiufeng generation directly without hand-writing HTTP.

The MCP Server is currently planned / coming soon. Once it ships, this page will list the official server name, installation, and tool inventory. Until then, use the OpenAI-compatible approach above.

Previous

Errors

On this page