MCP with custom config

Nobody has claimed this yet.

Tested with robocop, version 8.3.2

I've tested the robocop-mcp and it does work quite nicely with GitHub Copilot in VS Code.

In our RF development we have few custom rules and formatters and I'm not sure if such a configuration has been given a though how to set it up.

We have a separate project where our rule and formatter code lives.
Then running robocop check and format through that package we can use those custom rules and formatters with robocop runs using custom config.

Running MCP is a bit different in that how it can be installed, executed and configured from our package so that it has access to our custom code.
To install it and execute it with tool like `pipx` there's a custom entry point defined to run a thin wrapper which runs the MCP.
Then agent needs to provide the `config_path` to each call it makes to be able to take those custom configs into use (Unless I have missed some other way).

In our case,
as custom config is defined in the target project and "our-robocop-mcp" is configured per project it would be nice to set the default config file that would be used in case Agent is not providing it.
It is very difficult to know if and when the agent/LLM sets it (even if asked to). It is quite unfortunate if AI is formatting/creating lot of the code a wrong way it.

Maybe some environment variable that would set the default config file nothing provide by the agent call?

Locate the MCP entry point and trace how config_path is handled for each tool call. Read the configuration and wrapper setup before choosing how a project-level default should be supplied. Done means the MCP uses the configured default when an agent omits config_path, while an explicitly provided path still works.

Written by the indexing model from the issue text.