Comfyui MCP Server

Text prompts, saved graphs, and previously generated assets provide starting points for ComfyUI renders through comfyui-mcp. Its tools can construct template-based workflows, change node inputs or connections, and check for missing models or invalid wiring before submission. Mermaid conversion exposes pipeline structure, while queue controls track jobs and allow pending work to be edited or reordered. Parameter sweeps group variations into batches; image retrieval, execution history, and asset metadata support inspection and further iterations.

Model downloads, installed-model inventories, and custom-node installation support preparing the connected ComfyUI environment. GPU statistics and cached-model unloading address memory management, while diagnostics identify failed nodes and missing dependencies. Local provenance locks record model hashes, node-pack revisions, and the ComfyUI version for later comparison. The MCP server can connect to local or remote ComfyUI, with community support for Comfy Cloud as well. RunPod integration can manage GPU pods and switch rendering between a pod and the local machine. The issue-reporting tool describes the project as unmaintained and points users toward official ComfyUI tooling.