NamiFusion × your AI agent
Your agent already understands the project. Give it the tools to create the visuals—with NamiFusion MCP.
Example agent workflow
“Create a cinematic image for my next project.”
Try these starter prompts after connecting. Model availability and generation costs vary.
01 / Creative assets
Stay with your code or creative brief while your agent finds a model and creates an asset.
“Use NamiFusion to find an image model and generate a 16:9 editorial illustration of a quiet coastal town at sunrise. Return the output URL and the model used.”
Generation MCP
02 / From still to motion
Give your agent an accessible image URL and a motion brief. It can inspect model inputs and track the task.
“Find a NamiFusion image-to-video model. Use my image URL as the reference, add a slow camera push-in, and use a duration the model supports. Run one video and return the result.”
Marketplace MCP
03 / Connected production
Use the same saved canvas for visual planning and agent-assisted changes, with explicit node dependencies.
“Inspect my saved NamiFusion workflow WORKFLOW_ID. Summarize its assets and dependencies, then prepare an image-to-video branch using the available generation forms. Return the updated canvas cursor before generation.”
Workflow MCP
One account. Three connections.
Turn a prompt into images, video, or audio. Submit face swaps, follow task progress, and retrieve the finished output.
https://www.namifusion.com/api/v1/task/mcp/Search the model catalog, inspect each model’s inputs, and call its API through the same agent conversation.
https://www.namifusion.com/api/v1/marketplace/mcp/Read and edit your saved canvas, connect assets to image and video nodes, and generate selected nodes or groups.
https://www.namifusion.com/api/v1/workflow/mcp/A few lines. A new capability.
Sign in to NamiFusion and create a dedicated key on the API Keys page.
Choose your client and service. Copy the setup, replace the placeholder locally, and run it in your terminal.
Open your agent, check that NamiFusion tools are available, and try a starter prompt.
Have another MCP client? The integration guide includes HTTP authentication, Gemini CLI, and a local bridge option.
Terminal · Claude Code
export NAMIFUSION_API_KEY='sk-YOUR_API_KEY'
claude mcp add --transport http namifusion-generation \
https://www.namifusion.com/api/v1/task/mcp/ \
--header "Authorization: Bearer $NAMIFUSION_API_KEY"Replace sk-YOUR_API_KEY in your own terminal. Keep your key private.
Start here for image and video generation. Your agent discovers available models before submitting a task.
Claude Code saves the configured header in its local MCP configuration. Repeat for any other service you need. Full integration guide ↗
Before you connect
They can discover models, generate images, videos and audio, run face-swap tasks, retrieve results, and read or edit saved workflow canvases. Connect the Generation, Marketplace, or Workflow service for the capabilities you need.
No. NamiFusion provides remote MCP servers over Streamable HTTP. Create a NamiFusion API key, add a service URL to your client, and send the key in a Bearer authentication header.
The setup on this page is for Claude Code and Codex. Other clients need support for Streamable HTTP with a custom authentication header, or a compatible local bridge. A Claude desktop or ChatGPT connector that requires an OAuth flow is not covered by this API-key setup.
Generation and model runs consume NamiFusion credits according to the selected model and inputs. Usage and model charges still apply when you connect an agent. Check current model pricing and your account balance before running tasks.
Yes, when the client can upload files over HTTP. Generation MCP provides a signed upload URL. Workflow MCP provides a multipart upload flow that creates a durable asset. The client transfers the file bytes directly to storage, then references the uploaded media in its tool calls.
Start with a saved workflow, or create one with workflow_create. The agent reads saved canvas state and uses the current cursor when making changes. Workflow generation requires an API key without a model allow-list or monthly quota; workflow calls do not increment per-key usage counters. The integration guide explains the authoring and generation flow.
Explore the models, the canvas, and the ways to connect.