Github with PR Template Tools

The following illustrates basic GitHub MCP server that provides tools for analyzing git changes and suggesting appropriate PR templates.

Setup

1. Install uv

Follow the official installation instructions at: https://docs.astral.sh/uv/getting-started/installation/

2. Install dependencies

3. Configure the MCP Server

Add the server to Claude Code:

Related MCP server: Git MCP Server

Tools Available

1. **analyze_file_changes**: Analyzes Git changes between a base branch and HEAD to extract file-level modifications, commit history, and optionally truncated diffs — enabling automated PR generation, impact assessment, or review workflows.

• Key Functionalities:

  • Parses changed files via git diff --name-status into structured JSON: [{ "status": "M", "file": "main.py" }, ...]

  • Fetches change statistics via git diff --stat

  • Captures commit history using git log --oneline

  • Optionally includes full or truncated diff (default: 500 lines)

  • Smart fallback logic:

  • Uses mcp.get_context().session.list_roots() to infer working directory

  • Defaults to os.getcwd() if context is unavailable

  • Embeds a rich _debug block with root info, working directory trace, and server context

2. **get_pr_template**: Fetches the available PR templates from the shared template directory, including content and metadata — ready to be displayed, edited, or autofilled by Claude or similar agents.

• Key Functionalities:

  • Loads all templates defined in DEFAULT_TEMPLATES (e.g., bug.md, feature.md, docs.md, etc.)

  • Reads the content of each template file from disk

  • Fails gracefully if a file is missing or unreadable (Error loading template...)

  • Returns metadata alongside the file body: filename, type, content

3. **suggest_templates**: Provides Claude-friendly PR template recommendations based on a natural language change summary and change type (e.g., "bug", "feature"). Suggests the best-fit template and alternatives with reasoning and confidence level.

• Key Functionalities:

  • Normalizes change_type using a semantic map (TYPE_MAPPING)

  • Selects a recommended template (or defaults to feature.md)

  • Includes 2–3 alternatives to support Claude in ambiguous scenarios

  • Adds reasoning and confidence level ("high"/"medium")

  • Embeds the full template content to support AI-based autofill or mutation

• Inputs:

• changes_summary: natural language summary of the diff

• change_type: one of bug, feature, docs, refactor, performance, etc.

Future Improvements:

• Expose this as a web tool or MCP extension

• Embed automatic PR generation directly into GitHub Actions

• Integrate template generation with LLM-assisted refactoring

Author

Udayan Sawant | Generative AI Engineer

Building intelligent tools for the future of developer productivity