Building a read-only REW Pro evidence/audit workflow — looking for feedback

Hi, this is my first post. I started with a UMIK and REW a few months ago, and since then I started building a REW API / Claude MCP tool.

I understand that AI and audio do not go together for many people, and I get the scepticism. I am trying to use AI in a controlled way: not to make things up or automatically “fix” the room, but to help read measurements, check evidence, compare data, and avoid missing important details.

If some more experienced REW users could give me a clue whether this is something worth continuing, or whether I should just throw it in the bin, I would really appreciate it.

The first text is a public/forum-style explanation of the project: a local REW Pro MCP analysis workflow that lets an AI assistant inspect REW measurements through structured data instead of only screenshots. It explains the goal, limits, and intended read-only use.

The second text is a more technical workflow report. It describes how the tool should actually perform an analysis: preflight, measurement identity checks, fingerprinting, evidence reconcile, evidence audit, compare intent, interpretation, and report-pack export.

The tool can already do early versions of:
- measurement fingerprinting
- evidence/registry reconcile
- structured measurement comparison with evidence labels
- compact report-pack export with UUIDs, metadata, fingerprints, notes, and integrity hash
- separating raw evidence from derived interpretation

It is not meant to be automatic room correction, and it does not replace REW knowledge or acoustic judgement.

I would like feedback on:
- what is good about this approach
- what is risky or overcomplicated
- what evidence checks are missing
- what should remain manual
- whether this would be useful for serious REW users
- whether the workflow is clear enough for others to trust or criticize
```

Are you perhaps reinventing the wheel? Or is this your work? https://lobehub.com/mcp/koltyj-rew-mcp...

In order for everyone to use this tool, you must take into account 1000 variables, many of which cannot be explained in technical language.

​

sm52 said:

In order for everyone to use this tool, you must take into account 1000 variables, many of which cannot be explained in technical language.

Click to expand...

can u put here a complete REW sweep session .mdat file of your set up without giving me details ? then i give you the results off the analysis. would really apreciate this.

ddude003 said:

Are you perhaps reinventing the wheel? Or is this your work? https://lobehub.com/mcp/koltyj-rew-mcp...

Click to expand...

My apologies, I didn’t know someone had already made a friendly user MCP tool. The one I’m working on is for my own use, and I was planning to make it user-friendly once I know it works. Would you be so kind as to share a .mdat file without any extra info so I can show you the result my tool produces? It’s really just for testing, and you can give me your honest opinion on whether it’s terrible or usable—any feedback, good or bad, would be appreciated. Thanks in advance.

No worries and no need to apologize... If you use this website search tool and search post content + keywords = .mdat you could find many examples to test with... And then I would perhaps do some due diligence against several other AIs... Most of them seem to know REW...

Give it a try.

sm52 said:

Give it a try.

Click to expand...

hi, sorry that i didnt respond earlier in attachment u can find the result.

I read it. Your program extracts a lot of information from measurement files. Everything seems to be true. The only thing that doesn't match is the line - `mid shield` phase is unwrapped, reaching about +17,236 degrees. In fact, it's 1350 degrees. You did not say that the required measurements will be compared with each other. And the fact that the program evaluates the equalizer settings. I didn't need either of those. For an amateur, so much information will frighten him, so it is unlikely that this amateur will want to understand it. For a person who understands how to read measurements, such information is not needed. To compare the two measurements, there are All SPL and Overlay plots. Looking at them, the differences are immediately visible. Here's my opinion. I hope someone else will be interested and give their comments.

sm52 said:

I read it. Your program extracts a lot of information from measurement files. Everything seems to be true. The only thing that doesn't match is the line - `mid shield` phase is unwrapped, reaching about +17,236 degrees. In fact, it's 1350 degrees. You did not say that the required measurements will be compared with each other. And the fact that the program evaluates the equalizer settings. I didn't need either of those. For an amateur, so much information will frighten him, so it is unlikely that this amateur will want to understand it. For a person who understands how to read measurements, such information is not needed. To compare the two measurements, there are All SPL and Overlay plots. Looking at them, the differences are immediately visible. Here's my opinion. I hope someone else will be interested and give their comments.

Click to expand...

sm52 said:

I read it. Your program extracts a lot of information from measurement files. Everything seems to be true. The only thing that doesn't match is the line - `mid shield` phase is unwrapped, reaching about +17,236 degrees. In fact, it's 1350 degrees. You did not say that the required measurements will be compared with each other. And the fact that the program evaluates the equalizer settings. I didn't need either of those. For an amateur, so much information will frighten him, so it is unlikely that this amateur will want to understand it. For a person who understands how to read measurements, such information is not needed. To compare the two measurements, there are All SPL and Overlay plots. Looking at them, the differences are immediately visible. Here's my opinion. I hope someone else will be interested and give their comments.

Click to expand...

Thanks, that is useful feedback.

You are right that the report is too heavy for a casual REW user. I should probably split the output into two modes: a short comparison summary and a deeper forensic metadata report.

Also thanks for catching the phase range issue. If the correct value is about 1350 degrees rather than 17,236 degrees, then my phase-summary extraction is clearly over-reporting or reading the representation incorrectly. I will treat that as a bug/check item rather than a conclusion.

The tool is not meant to replace All SPL or Overlay plots. Those are still the fastest way to visually compare two traces. The purpose of my tool is more to produce a reproducible text audit: what was compared, what metadata differs, which claims are safe, and which claims should be blocked.

For shared forum files I agree the output should be much shorter: basic metadata, FR differences, timing/calibration caveats, and a short conclusion only.

I also just posted another demo using the official REW sample .mdat file. That one may be a better example of what I am trying to test, because it includes mixed SPL and impedance data and shows how the tool separates domains and blocks invalid comparisons. If you have time, I would appreciate your feedback on that one too.

I am not free to engage in activities distracted from my own tasks for a long time. Perhaps someone else will be interested in this.