Work Intelligence Agent

Automated weekly work reporting from git history across multiple repositories — now with optional AI-powered summaries via Google Gemini.

What It Does

Point the agent at a directory (or list of repos) and an author name, and it produces a clean markdown report of the week's work — grouped by day, organized by repository, with similar commits summarized together.

New in v0.2.0: When a Gemini API key is configured, groups of 3+ related commits get an AI-generated one-line summary for faster reading.

Installation

Requires Python 3.11+ and git on your PATH.

pip install work-intelligence-agent

Or run directly with uvx:

uvx work-intelligence-agent --author "Your Name" --parent-directory ~/projects

AI Summarization (Optional)

Set your Gemini API key to enable AI-powered commit summaries:

export GEMINI_API_KEY="your-gemini-api-key"

Get a free API key at Google AI Studio.

Without the key, the agent still works — it just uses prefix-based grouping instead of AI summaries.

CLI Usage

# Scan a parent directory for repos and generate a report
work-intelligence --author "Jane Doe" --parent-directory ~/projects

# Specify repos explicitly
work-intelligence --author "Jane Doe" --repositories ~/projects/repo-a ~/projects/repo-b

# Custom date range
work-intelligence --author "Jane Doe" --parent-directory ~/projects \
  --start-date 2025-01-06 --end-date 2025-01-12

# Write output to a file
work-intelligence --author "Jane Doe" --parent-directory ~/projects --output report.md

CLI Options

Configuration File

Place a JSON config at .kiro/work-intelligence.json to set defaults:

{
  "author": "Jane Doe",
  "parent_directory": "~/projects",
  "date_range": {
    "start": "2025-01-06",
    "end": "2025-01-12"
  }
}

CLI arguments override config file values.

MCP Server

The agent also runs as an MCP server for integration with AI tools like Kiro.

{
  "mcpServers": {
    "work-intelligence-agent": {
      "command": "uvx",
      "args": ["--from", "work-intelligence-agent", "work-intelligence-mcp"],
      "env": {
        "GEMINI_API_KEY": "your-gemini-api-key"
      }
    }
  }
}

When used via MCP, the agent will prompt you for any missing details (author name, repo location, date range) through the AI assistant conversation.

The server exposes three tools:

• generate_report — Full weekly report generation (prompts for author, repos, dates, and output path)
• scan_repos — Discover git repos under a directory
• extract_history — Extract raw commit history

Development

# Install with test dependencies
pip install -e ".[test]"

# Run tests
pytest

# Run the MCP server locally
work-intelligence-mcp

Changelog

v0.2.0

• AI-powered commit summarization via Google Gemini (gemini-2.0-flash)
• Set GEMINI_API_KEY to enable — falls back gracefully without it
• MCP server now prompts users for missing parameters (author, repos, dates)
• generate_report tool now accepts output_path to save reports to file
• Added google-genai dependency

v0.1.0

• Initial release
• CLI and MCP server for weekly work report generation
• Prefix-based commit grouping and summarization
• Recursive git repository scanning

License

MIT