How to run like the original?

[rohan0rtiz0]

I followed ClearBlueJar's walkthrough here and I was able to use the original GhidraMCP with a local LLM, which is a must for my work.

After getting things glued to the OpenWeb-UI (which was basically just a docker command), the meat of running GhidraMCP was running uvx mcpo --host localhost --port 1337 -- uv run bridge_mcp_ghidra.py.

I wanted to try out your improvements to GhidraMCP to see if they yielded improved results for my analysis-acceleration tests, but it seems like you don't have a bridge_mcp_ghidra.py anymore like the original (although there's still vestiges of it in the documentation). And the current install suggestion of npx -y @smithery/cli install @DaCodeChick/GhidraMCP --client claude mentions claude and therefore it seems like it wouldn't be the offline version I need.

Do you have any suggestion for what the minimum change would be to update the steps from the ClearBlueJar walkthrough to work with your plugin fork? Thanks a bunch.

[leumasme]

Sadly doesn't seem like this project is currently usable. The python scripts in the latest release are very broken. The current git head changes them majorly to maybe work, but here the actual ghidra plugin java doesn't even build due to a dozen syntax errors.