Midjourney opens a limited MCP test so AI agents can call its models
Midjourney is picking a small group to test an official MCP server that lets AI assistants call its models, for creative work only and not inside SaaS products. It has published no tool list, limits, price or timeline.
Midjourney wants to let AI agents drive its models. In a post on Oct 10 it said it will test a Midjourney MCP with a limited community, and linked an application form. MCP, short for Model Context Protocol, is the plug-in standard that lets an assistant such as Claude or ChatGPT call an outside tool, so an agent could prompt Midjourney inside a larger job instead of you pasting prompts across.
The scope is narrow by design. Midjourney says the test is "not for Midjourney in SaaS products" and is for "purely artistic and creative endeavors." It has published no tool list, no rate limits, no price and no timeline.
The form says more than the post. It asks which assistant you would connect (Claude, ChatGPT, GLM, DeepSeek, Kimi, Mistral, Qwen or Cursor), how many images a week you expect to make, from 1 to 10 up to more than 1,000, and the most you would pay for access later, with options from $10 to $1,000 a month that include baseline credits.
In practice, if you storyboard or build look development with an agent, treat this as a selection round rather than a launch: Midjourney says your answer on what you want to make is important for picking testers, so describe one specific project and link your past work.
Source: Midjourney on X and its MCP application form, Oct 10, 2026.
