FlutterFlow AI CLI validate/run crashes on macos_x64 / Intel Mac after MCP read access succeeds

[franzzedler]

Can we access your project?

• I give permission for members of the FlutterFlow team to access and test my project for the sole purpose of investigating this issue.

Current Behavior

FlutterFlow AI CLI / MCP read operations work successfully on an Intel MacBook Pro 2019. status, inspect, and resources all succeed and Codex can inspect the project.

However, when attempting to validate or run a tiny text-only edit, dart test / flutterflow ai validate crashes with a Dart VM error / Abort trap: 6.

flutterflow ai doctor reports:

Validator runtime: macos_x64
Validator platform support: unsupported
Supported platforms: macos_arm64, linux_x64, linux_arm64
Validator download is unavailable on this runtime

Expected Behavior

Either:

• macos_x64 should be supported by the FlutterFlow AI validator,
  or
• the CLI should fail gracefully with a clear recommendation for Intel Mac users, for example “use linux_x64 / Codespaces for validated edits,” and avoid a Dart VM crash.

Steps to Reproduce

1. Use an Intel MacBook Pro 2019 / macos_x64.
2. Set up FlutterFlow AI CLI workspace.
3. Confirm that flutterflow ai status , inspect, and resources work.
4. Use Codex MCP or CLI workflow to prepare a tiny text-only edit.
5. Run dart test or flutterflow ai validate dsl/edit.dart --project-id .
6. Observe Dart VM crash / Abort trap: 6.

Reproducible from Blank

• The steps to reproduce above start from a blank project.

Bug Report Code (Required)

not inside FF

Visual documentation

Environment

- MacBook Pro 2019, Intel
- Runtime: macos_x64
- FlutterFlow AI Doctor version: 0.2.0
- Dart SDK: 3.11.5 stable
- MCP read access succeeds
- Validation/edit pipeline fails locally

Additional Information

This blocks safe MCP editing for Intel Mac users, even though read-only MCP access works. It would be very helpful for early-stage founders, students, educators, and solo builders if macos_x64 support or a documented cloud/Linux workaround were available.

[vairagadeharshwardhan]

On windows mcp dsl push says the file size is capped at 64 KB! WHY IS THAT?

[Alezanello]

Hello,

The engineering team has confirmed that this issue was resolved in version 0.0.34. Please update to the latest version and let me know if everything is working as expected.

[franzzedler]

@Alezanello Update after testing 0.0.35:

I upgraded from flutterflow_cli 0.0.34 to 0.0.35:

dart pub global list | grep flutterflow_cli
→ flutterflow_cli 0.0.35

I then ran:

flutterflow ai upgrade

After that, the situation improved significantly on my Intel MacBook Pro 2019 / macos_x64.

Previously I was seeing crashes / validation failures. With 0.0.35, I was able to create a small local dsl/edit.dart from Codex and apply a tiny text-only change using path-based selectors.

So from my side, 0.0.35 seems to have resolved the main blocking issue for this small test case on Intel Mac.

One important question:
I ran:

flutterflow ai validate dsl/edit.dart --project-id ...

The output showed:

[OK] compileDslApp
Pushed project ... (commit ...)

Can you clarify whether flutterflow ai validate is expected to push changes to the FlutterFlow project? I expected validation to be dry-run/no-push, so this distinction is important for safely using MCP on production projects.

Thank you — this is a big improvement.

[franzzedler]

One nuance: flutterflow ai doctor still reports the local validator runtime as macos_x64 / unsupported, but the actual flutterflow ai validate ... command no longer crashes for me. I was able to compile and apply a tiny path-based text fix from Codex, and the change appeared correctly in FlutterFlow.

So it looks like the remote validator/fallback is now working for this small test case. Thank you! 🙏

So, as stated above, only one question remains: my validate command output said Pushed project ... (commit ...). Is validate expected to push, or should it be dry-run only?