MCP & AI agents

The current Python MCP server helps coding agents inspect SDK capabilities and author or validate portable visual workflows from installed node schemas. It does not automatically execute graphs, access the browser's live WSI, or control the viewer.

Start with MCP setup and workflow authoring. The local workflow guide covers execution in the viewer and companion. Keep these two roles distinct: MCP authoring produces workflow definitions; the viewer and local Python companion run them after explicit pairing.

Current limits

Agent capabilities depend on the tools and schemas exposed by the installed SDK. This site does not imply autonomous model execution, browser control, remote slide access, or patient-data upload by an MCP client.