FluxMCP

FluxMCP is a Model Context Protocol (MCP) server for AI image generation and editing using Flux models through the AceDataCloud platform. Users can generate and edit images directly from Claude, Cursor, or any MCP-compatible client.

Key capabilities:

• Image generation: Generate images from text prompts using 6 Flux models (including flux-dev, flux-pro, flux-kontext).
• Image editing: Edit existing images using the context-aware Flux Kontext model.
• Task management: Track asynchronous generation tasks and support batch status queries.
• Model guidance: Built-in guidance for model selection and prompt writing.
• Dual transport modes: Supports stdio (local) and HTTP (remote/cloud) modes.
• Docker ready: Provides containerization and K8s deployment manifests.
• Security: Uses Bearer token authentication in HTTP mode and implements per-request isolation.

Usage: Users can connect via hosted OAuth, API token, or local stdio. The hosted endpoint is https://flux.mcp.acedata.cloud/mcp。托管服务支持 OAuth metadata and Dynamic Client Registration (DCR); DCR registers client applications, not API keys. After OAuth login, the client sends a Bearer token and may need to reuse or create the account's API credentials. Browser login still requires an AceDataCloud account. The hosted service may be billed; before actual operation, check the current service documentation and displayed pricing. Do not configure both OAuth login and a fixed Authorization header for the same server.

Configuration examples for each client:

• Claude and Claude Desktop chat: Add a remote custom connector in Customize → Connectors → Add custom connector, enter https://flux.mcp.acedata.cloud/mcp，选择登录，若 If Claude asks how to register the OAuth client, choose automatic registration and complete consent. Claude Desktop's local claude_desktop_config.json is a separate setting.
• Claude Code: Run claude mcp add --transport http --scope user flux https://flux.mcp.acedata.cloud/mcp，然后 claude mcp login flux. Check /mcp.
• Cursor: Add a remote server containing only https://flux.mcp.acedata.cloud/mcp . Project configuration is merged into <project>/.cursor/mcp.json; for personal use, use ~/.cursor/mcp.json.
• VS Code / Copilot: Run MCP: Add Server, select HTTP, enter https://flux.mcp.acedata.cloud/mcp，然后完成浏览器登录。新的可移植工作区配置使用 <project>/.mcp.json; for the VS Code-specific format, use <project>/.vscode/mcp.json or the user configuration file. Check MCP: List Servers.
• Codex: Run codex mcp add flux --url https://flux.mcp.acedata.cloud/mcp，然后 codex mcp login flux. User settings are located at ~/.codex/config.toml.

Managed API token method: Open Ace Data Cloud → Applications, log in or register an account. Find the General application; if it has not been created, choose Initialize General Application. A general key can use the services available to the account; a dedicated application's key is limited to its services. Check the required service/model access permissions and current pricing. Copy the API Key from the application card, or open Keys / Manage Keys → Create to create a separate key. When creating a key, review Expiration Time, Set Usage Limitation, and Allowed APIs, enable only the required limits and the operations allowed to be called, then click Create and copy the token from the key list. Task queries do not grant permission to submit generations. Confirm that the application balance can cover the first request. Copy only the token value, without quotes, spaces, or Bearer. The API Key / API Token field expects the raw value; only a manually configured HTTP Authorization header needs the Bearer prefix. A platform-... account management token cannot replace a generated API key. Store the key in the client credential store or an environment variable. Never put it in prompts, shared workflows, screenshots, or Git commits. If authentication fails, check the complete copied value, application/service access, expiration time, usage limits, and balance before creating a new key.

Local installation: Install the package and provide the API token to the local process. For Claude Desktop local MCP, merge this entry into the file opened by developer settings (on macOS, ~/Library/Application Support/Claude/claude_desktop_config.json). uvx requires uv to be in PATH. Keep this user-level file private. For self-hosted HTTP, use mcp-flux-pro --transport http --port 8000; expose it only through appropriate network and TLS controls. Local execution still calls AceDataCloud API.

Pre-use checks:

• https://flux.mcp.acedata.cloud/health Returning {"status":"ok"} only checks endpoint reachability.
• Confirm that the MCP client loads the tools. flux_list_models is a reference tool and does not verify downstream API access or balance.
• If a full API check is needed, after reviewing the current service documentation and displayed pricing, call flux_generate_image with your own valid input. If the result contains a task ID, call flux_get_task for the same ID until it finally succeeds or fails. Do not resubmit the operation just to check progress.
• 401 error: check the authentication route used by the client and whether the token or OAuth session is valid. 403 may indicate an account permission or content moderation failure; read the returned error. Insufficient balance and downstream service failures require their own diagnostics. Listed tools or submitted tasks do not prove a successful result.

Available tools and prompts: Provides available tools and available prompts, supporting image generation and task status checks.

Supported models: Supports 6 Flux models including flux-dev, flux-pro, flux-kontext, and others.

Environment variables: This MCP server uses the AceDataCloud Flux API:

• POST /flux/images — generate or edit images
• POST /flux/tasks — query task status (single or batch) Full API documentation: platform.acedata.cloud

Video generation: Use flux_generate_video to support t2v, i2v, v2v, or draft_enhance with a structured request. For example: {"mode":"t2v","prompt":"Waves at sunset","duration":5,"resolution":"hd","generate_audio":false}. Image modes require keyframes; video modes require start_video. Draft enhancement uses an owned platform draft_task_id and requires that the temporary draft cache is still available. Video generation uses POST /flux/videos, with action=generate (default) and mode=t2v/i2v/v2v/draft_enhance. Each generation mode accepts its own fields. The tool returns a task ID asynchronously by default; use flux_get_task retrieves the final video. Set async=false in the request to wait synchronously.

License: MIT License — see LICENSE for details. Related: AceDataCloud Platform, Flux by Black Forest Labs.