At SIGGRAPH 2026, Adobe, Blender, Unreal Engine, Houdini, Canva’s Affinity suite, and Boris FX Silhouette announced MCP server support. Model Context Protocol is an open standard through which AI systems can use tools exposed by software. What an agent can do depends on the particular server, its APIs, and the permissions it receives.
Check what each integration exposes
Adobe shipped an MCP server for Express so AI coding assistants can build add-ons using official APIs. Blender launched Blender Lab, described as a lightweight server exposing its Python API to agents through natural language. Unreal’s integration lets agents reason over scenes and project state in real time; Houdini connects MCP to its APEX Script workflow.
Boris FX Silhouette exposes operations for inspecting projects, building node trees, editing keyframes, and rendering frames. Canva brought its Affinity suite into the mix with an AI Connector for Claude for natural-language automation in the design workspace.
These integrations cover different parts of production. A team needs to check the supported operations in its chosen tool before planning a workflow around them.
Give the agent a production assignment
Consider a hypothetical campaign workflow: an agent reads a brief, prepares layout variations in Express, and creates social templates in the required formats. A designer reviews the files and decides what to develop further. Each of those steps needs a supported tool operation and a clear handoff; MCP access by itself does not establish that the whole sequence will run reliably.
The same approach applies to a Houdini scene or Unreal project. Give the agent a bounded assignment, such as inspecting a node tree or identifying assets for review, and check the result inside the project. The creative lead still decides whether the proposed change fits the work.
Try it on a copy of the project
Pick a repeatable task such as asset resizing, template generation, or scene setup. Check the integration’s availability and permissions, then run it on a copy of the working files with a defined reviewer. Record what it produced, what needed correction, and whether the handoff saved the team time.
That small test will tell you more about the integration’s place in your production process than the breadth of the announcement list.