Can an MCP tool help AI compare photos with CAD drawings and flag mismatches?
Special_Spring4602 · reddit · 2026-07-24
The poster is exploring whether an AI can reliably compare a photo of a built part against its CAD drawing and flag mismatches.
The proposed approach is an MCP tool that would:
- read DWG/DXF files and extract the drafter’s annotated dimensions instead of guessing from line geometry,
- pass exact coordinates and measurements to the model,
- use the model for vision on the photo, while a separate tool handles pass/fail tolerance math.
Before building it, the author asks whether this already exists, whether file parsing is useful compared with live AutoCAD/COM access, and whether the real bottleneck is actually getting a usable photo rather than the comparison step itself.
More from coding & agent
- Pi 0.82.0 lets bash commands detect the loaded model and provider — mitsuhiko · 2026-07-24
- SGLang talk says RL post-training is really an inference problem — BanghuaZ · 2026-07-24
- Claude’s code-comment narration looks like the same failure mode as Bash-comment reasoning — mertdumenci · 2026-07-24
- Engineer says U.S. models block harmless security tests, forcing a switch to Chinese AI — kristoph · 2026-07-24
- Agent platform builders debate whether custom prompts belong in system or user messages — Final-Choice8412 · 2026-07-24
- rem-exec wraps SSH commands in JSON for agent-friendly remote operations — nivosnation · 2026-07-24