Robocop MCP server

Robocop MCP server helps users to resolve their static code analysis errors and warnings with help of an LLM. It has two tools to help resolve Robocop rules:

Get robocop report

Get robocop report tool allows calls robocop check command and returns one rule violation (maximum 20 from one rule) to LLM. The rule violation summary also contains default recommendation how the rule violation can be fixed. User than can accept the suggested proposal or tell LLM a different way to fix the rule violation.

The maximum number of rule violations, proposed fix for a rule and many more things can be configured in the pyproject.toml file. See Configuration chapter for more details.

Run robocop format

You can also run robocop format tool. Because robocop has complex commandline syntax, robocop-mcp only support giving file or folder as an argument for the format command. Rest of the configuration should be placed in the robocop configuration file. The --reruns option is set to 10.

Install

Install with pip: pip install robocop-mcp

Running robocop-mcp server

running MCP server in VS Code workspace:

1. Create a .vscode/mcp.json file in your workspace.
2. Add following configuration to the mcp.json file:

{
    "servers": {
        "robocop-mcp":{
            "type": "stdio",
            "command": "${workspaceFolder}/.venv/bin/python",
            "args": [
                "-m",
                "robocop_mcp",
            ],

        }
    }
}

3. Change your CopPilot chat to Agent mode and select suitable model for your use.
4. Remember to click start button in the mcp.json file

For general detail about configuring MCP server in VS Code, see the VS Code documentation

Using robocop-mcp

robocop_mcp_announcement.mp4

Configuration

The robocop-mcp server can configured by using pyproject.toml file. The robocop-mcp server uses [tool.robocop_mcp] section in the toml file.

To robocop-mcp server see the the toml file, a ROBOCOPMCP_CONFIG_FILE environment variable must be set. Example in the mcp.json file:

{
    "servers": {
        "robocop-mcp":{
            "type": "stdio",
            "command": "python",
            "args": [
                "-m",
                "robocop_mcp",
            ],
            "env": {"ROBOCOPMCP_CONFIG_FILE": "${workspaceFolder}/pyproject.toml"},
        }
    }
}

Priority of Robocop rules

Some rules are more important to fix than others or perhaps you want to use certain type of LLM to solve certain type of rule violations. In this case you can use rule_priority (list) to define which rule are first selected by the robocop-mcp and given to the LLM model. The rule_priority is a list of robocop rule id's. You can list all the rules with command:

> robocop list rules

And if one one rules looks like this:

Rule - ARG01 [W]: unused-argument: Keyword argument '{name}' is not used (enabled)

Then rule id is the ARG01.

And example if user wants to prioritize the ARG01 and ARG02 to be fixed first, then rule_priority would look like this.

[tool.robocop_mcp]
rule_priority = [
    "ARG01",
    "ARG02"
]

It is also possible to define rule priority by rule name. Example if there is a need to define ARG01 and ARG02 rules by name, then rule_priority would look like:

[tool.robocop_mcp]
rule_priority = [
    "unused-argument",
    "argument-overwritten-before-usage"
]

If rule_priority is not defined, robocop-mcp will select take first rule returned by robocop and use it to find similar rule violations. If no rules match to rule_priority list, first rule returned by Robocop is used.

Maximum amount violations returned

To not to clutter the LLM context with all the rule violations found from the test data, by default robocop-mcp will return twenty (20) violations from robocop. This can be changed by defining different value in the violation_count (int) setting.

To make robocop-mcp return 30 rule violations:

[tool.robocop_mcp]
violation_count = 30

How many rule violations the robocop-mcp should return depends on the LLM model being used, how verbose the proposed fix is and how long the LLM model context have been in use. It is hard to give good guidance on this subject, because LLM models change at fast pace and there are some many different models available.

Custom fix proposals

Each rule violation contains robocop default rule documentation how the problem can be addressed. In some cases, this may lead to LLM to wrong solution or you want to apply custom way to fix the specific rule. Custom solution can be defined in text file (markdown is recommended, because it is easy for LLM to understand.) and defining custom rule files in the pyproject.toml. Each rule where custom fix is defined is defined as key in toml file and value must point to a text file.

Example if there need to define custom fix for ARG01, create ARG01.md file, example in a my_rules folder. Then pyproject.toml should have:

[tool.robocop_mcp]
ARG01 = "my_rules/ARG01.md"

It is also possible to define custom fix proposals by rule name. Example to provide custom fix proposal for ARG01 by name, then toml file would look like:

[tool.robocop_mcp]
unused-argument = "my_rules/unused-argument.md"

Ignore rules

The recommended way to ignore rules is to ignore rules in the robocop tools section in the pyproject.toml. In that case rules are ignored by Robocop and ignored rules are not visible for the robocop-mcp serve either.

If there is need to ignore rules only for robocop-mcp, then add ignore (list) setting to the pyproject.toml file. Example if there is a need to ignore DOC02, DOC03 and COM04 rules, then pyproject.toml should have:

[tool.robocop_mcp]
ignore = ["DOC02", "DOC03", "COM04"]

It is also possible to to ignore rules by rule name. If same rules as in above would need to ignored, then ignore list would look like:

[tool.robocop_mcp]
ignore = [
    'missing-doc-test-case',
    'missing-doc-suite',
    'ignored-data'
]

Support separate robocop configuration file

Although robocop-mcp only supports pyproject.toml, the robocop itself does support multiple different configuration files. If your robocop configuration is not in pyproject.toml, then the separate configuration file can be defined in ROBOCOPMCP_ROBOCOP_CONFIG_FILE environment variable. Example mcp.json:

{
    "servers": {
        "robocop-mcp":{
            "type": "stdio",
            "command": "${workspaceFolder}/.venv/bin/python",
            "args": [
                "-m",
                "robocop_mcp",
            ],
            "env": {
                "ROBOCOPMCP_CONFIG_FILE": "${workspaceFolder}/pyproject.toml",
                "ROBOCOPMCP_ROBOCOP_CONFIG_FILE": "${workspaceFolder}/robocop.toml",
            }
        }
    }
}

Robocop reruns

It is possible to configure how robocop format --reruns command line argument is set. This is controlled in the pyproject.toml file by reruns (int) setting. By default value is set to ten, but the set it to two add the following to pyproject.toml file:

[tool.robocop]
reruns = 2

Contributing fix instructions for rule

Users can contribute their instructions to for rule fixes in the repository. This is explained in the CONTRIBUTING.md file. If there is not fix, either from robocop-mcp or by user defined fixes, then robocop documentation is used as fix instruction to LLM. User can always write their own fix explanation in the prompt, but if that is not one off, then it is easier to define user defined rule fix in a file or contribute a fix to the robocop-mcp project.