AI Agent Canvas — Visual Creative Workflows
Connect references, image generation and video steps, or describe a workflow for an agent to build.
Plan a multi-step creative task on the Framedance node canvas. Connect a brief or reference to image generation, then send an approved image into a video step. The in-canvas agent can help arrange a workflow from natural-language instructions, while the visual graph lets you inspect inputs and dependencies. Check model choices and costs before running generated steps.
Visible dependencies
See which prompt, reference and generated asset feeds the next operation.
Agent-assisted planning
Describe the intended output and review the proposed steps, models and connections.
Iterate selected stages
Review intermediate images and refine the relevant stage before spending credits on downstream video.
How to use
- 1Open a blank canvas or an available template and describe the deliverable, references and required stages.
- 2Add nodes or ask the agent to arrange them. Inspect connections, required inputs, selected models and costs.
- 3Run and review intermediate results. Revise the affected stage, then continue the workflow and export the approved assets.
FAQ
What does an AI video agent do here?+
It helps plan and connect creative steps such as reference preparation, image generation and video generation. You still review the workflow and outputs; available operations depend on the current tools and models.
Can I use the canvas without writing code?+
The visual canvas and natural-language agent support interactive authoring. Developers can also use the documented MCP integration where its tools match the workflow.
Will the entire workflow run without review?+
Review inputs, connections, model settings and cost before execution. Inspect intermediate outputs so an unsuitable image does not carry into subsequent paid steps.
Related guides
AI Video Agent Workflow: Plan and Run a Multi-Step Canvas
Use a canvas workflow and its MCP tools to organize assets, prompts, dependencies, execution, and review for a video project.
Image-to-Video Workflow: Turn a Still Into a Controlled Shot
Prepare a strong first frame, specify motion, choose compatible controls, and review an image-to-video result shot by shot.
Batch AI Generation API: Submit, Poll, and Validate Jobs Safely
Build a bounded batch runner for the Framedance marketplace API with schema discovery, rate handling, polling, and output validation.