Figma: Weave image and video models now accessible via MCP for external agents
In the week of October 26, 2026, Figma made the image and video generation models from Figma Weave — including Seedance, Kling, and Nano Banana — accessible to external AI agents through the Figma MCP server. Previously, Weave models could only be used from within Figma directly; this change means that agents running outside Figma, such as those in Claude Code or OpenAI Codex, can now call these models as part of a broader workflow.
What this changes for design and coding agents
Before this update, an external agent working in Figma through MCP could read and write design data but had no access to Weave’s media generation capabilities. Those required opening Figma itself and running models through the Weave interface. The update removes that break in the workflow: an agent that is already manipulating a Figma file can now generate images or video through the same connection.
Generations consume Weave credits, which are managed separately from Figma credits. Teams using external agents to automate design tasks will need to account for Weave credit usage in any agentic workflows that call these models.
Why it matters for product designers
The practical significance is that design and development workflows driven by external agents can now include AI-generated visual assets without requiring a human to switch contexts and run Weave manually. An agent building a screen from a prompt can pull generated imagery into the same file, using the same MCP connection. For teams already running Figma through an MCP-connected agent, this is an additive capability rather than a new infrastructure requirement.
Figma Weave launched in 2026 after Figma acquired Weavy, a node-based AI canvas for generating multimedia content. The models now available via MCP include both image and video models, with the exact list subject to change as Figma expands the Weave library.