SketchUp Agent Harness

SketchUp Agent Harness is an early-stage open-source project for connecting design agents such as Claude Code and Codex CLI to SketchUp. It is described as a harness for a verifiable design workflow, with the goal of letting designers express intent in natural language and turn that intent into executable, repairable project state.

The project addresses a gap in AI design tools that stop at image generation or a few software actions. Its core idea is that design work needs a structured source of truth. In this setup, design_model.json is the key file, while SketchUp scenes, screenshots, and renders are treated as execution results or review artifacts rather than the only long-term record. The system also uses bridge execution traces, component metadata, runtime skills, and protocol documentation.

Documented capabilities include initializing an isolated design project directory, installing the SketchUp Ruby bridge, installing Claude and Codex runtime skills into the project, generating simple spaces and bathroom layouts through natural language, importing DWG, DXF, PDF, image, scan, or photo inputs into a first editable model, and writing SketchUp entity IDs, execution status, and operation metadata back into project truth. The workflow is intended to support diff, validation, repair, and replay, and the project says that structured project state can preserve source evidence, screenshots, versions, project rules, imports, assets, and snapshots.

The project is aimed at designers and at people using agent CLIs in a SketchUp workflow. Claude and Codex are described as entry points, and the more stable core is said to be the MCP server, project workspace, design_model.json, bridge execution trace, SketchUp Ruby bridge, component metadata, runtime skills, and protocol documentation. It is delivered as an open-source project, with a public repository at github.com/marlinBian/sketchup-agent-harness. The page does not give licensing or pricing details beyond that open-source description.