MCPFlux

What AceDataCloud/MCPFlux MCP server does

AceDataCloud/MCPFlux MCP server connects MCP clients to Flux models available through the Ace Data Cloud platform. It supports two primary image workflows: generating an image from a text prompt and editing an existing image with text instructions. The server also exposes model and action discovery tools, so clients can inspect supported Flux models and available workflows rather than relying only on fixed configuration.

The repository identifies support for models including flux-dev, flux-pro, and flux-kontext, with the README describing six Flux models for generation. Editing uses context-aware Flux Kontext models. Image requests are handled as asynchronous tasks, allowing a client to submit work and check its progress or result later.

How it works

MCP-compatible clients call the server's tools. flux_generate_image accepts a text-based image request, while flux_edit_image applies a text instruction to an existing image. flux_get_task checks one task, and flux_get_tasks_batch checks multiple tasks in one request. flux_list_models returns model capabilities, and flux_list_actions describes available Flux actions and use cases.

AceDataCloud/MCPFlux MCP server supports local stdio transport and HTTP transport. The hosted endpoint is https://flux.mcp.acedata.cloud/mcp; hosted requests use a Bearer token. The local process can also run in HTTP mode, including inside a Docker container. HTTP mode supports per-request token handling, while the local stdio example supplies a token through the environment.

The project also provides MCP prompts for model and tool selection, prompt-writing guidance, and common workflow examples. These prompts are named flux_image_generation_guide, flux_prompt_writing_guide, and flux_workflow_examples.

Setup and configuration

For the hosted service, create an account on the Ace Data Cloud platform, acquire an API token, and send it as a Bearer credential. Claude.ai is documented as a separate OAuth connection path that does not require entering the API token. Claude Desktop, Cursor, Windsurf, Cline, and other clients can connect to the hosted streamable HTTP URL through their MCP configuration.

For local use, install the mcp-flux-pro package from PyPI and set ACEDATACLOUD_API_TOKEN. The stdio process is started with mcp-flux-pro; the README also shows uvx mcp-flux-pro. HTTP mode can be started with mcp-flux-pro --transport http --port 8000. Docker users can run the published image ghcr.io/acedatacloud/mcp-flux-pro:latest and expose port 8000.

Tools and capabilities

• Generate images from text prompts with model selection.
• Edit existing images using Flux text instructions.
• List Flux models and their capabilities.
• List available actions and workflow guidance.
• Query one asynchronous task or several tasks in a batch.
• Connect through local stdio, hosted HTTP, or self-hosted HTTP deployment.

Limitations and notes

An Ace Data Cloud API token is required for the documented hosted and local API workflows, except for the Claude.ai OAuth connection described in the README. The supplied material does not specify API pricing, quotas, image retention, or detailed input and output schemas. The available model table is incomplete in the provided excerpt, so model-specific limits should be checked through flux_list_models or the provider documentation. AceDataCloud/MCPFlux MCP server is an interface to Flux through Ace Data Cloud; it does not describe a standalone image-generation model hosted in the repository.