Commit Conventional Message Court (CCMCP)
# Commit Conventional Message Court MCP
Please write better commits with AI.
Give this repo to your AI and ask it to install the MCP for you.
Seriously. That is a valid setup flow.
I take your vague commit messages, drag them into court, point at the diff, and ask the question your teammates are too tired to ask:
"What exactly did you do?"
## Add to Codex
Add this to [`~/.codex/config.toml`](C:\Users\chrisdnm\.codex\config.toml):
```toml
[mcp_servers.commit-message-court]
command = "node"
args = ['C:\Users\chrisdnm\Desktop\commimWizard\dist\index.js']
```
Then restart Codex.
## Add to Claude Code
On Windows, add it as a local stdio server with:
```bash
claude mcp add --transport stdio commit-message-court -- node C:\Users\chrisdnm\Desktop\commimWizard\dist\index.js
```
If you want it available everywhere, add `--scope user`.
```bash
claude mcp add --transport stdio --scope user commit-message-court -- node C:\Users\chrisdnm\Desktop\commimWizard\dist\index.js
```
## Add to Claude Desktop
Add this to your `claude_desktop_config.json`:
```json
{
"mcpServers": {
"commit-message-court": {
"type": "stdio",
"command": "node",
"args": ["where code is"],
"env": {}
}
}
}
```
Then restart Claude Desktop.
## What This Thing Does
- It sees `fix stuff` and objects.
- It sees `final_final_real` and requests counsel.
- It sees a risky diff and asks for evidence.
- It sees your commit history and tries, against all odds, to make it readable.
-
## License
- [WTFPL](LICENSE) -- Do What The Fuck You Want To Public License.
TDQS
Scored across 3 tools
While each tool has a distinct function, the relationship between prosecute_commit and render_verdict is unclear—agents may not know if they need to call both in sequence or if prosecute_commit alone suffices. require_better_subject also feels like it could be a sub-operation of the prosecution rather than a standalone tool.
All tools follow a consistent court.<verb>_<object> pattern using lowercase snake_case. The 'court.' prefix acts as a namespace, and the verb choices (prosecute, render, require) are thematically appropriate, though 'require' breaks the pure verb_noun pattern slightly with the 'better' modifier.
Three tools is at the lower bound of acceptable for this domain. While the core analysis and suggestion functions are present, the set feels thin—missing configuration, batch processing, or actual commit amendment capabilities that would make this a complete solution.
The set covers analysis (prosecute), presentation (render), and suggestion (require), but lacks obvious lifecycle operations like applying fixes, configuring rules/standards, or handling commit metadata beyond the subject line. The separation between rendering and prosecution creates a potential gap if agents expect a single 'check and report' operation.