Agentic Swear Jar: A Local Dashboard for Analyzing Swear Words in AI Coding Sessions

Agentic Swear Jar is a private, local dashboard designed to analyze the usage of profanity within the session histories of your coding agents—Claude Code and Codex CLI. This tool provides valuable insights into the linguistic aspects of coding, transforming the frustration and challenges often encountered in software development into a quantifiable and entertaining report. By leveraging simple commands without uploading any prompts to an external API, Agentic Swear Jar enhances the experience of understanding coding sessions while keeping everything local and secure.

Key Features

• Local Analysis: No data is sent to external servers; the entire analysis takes place on your machine.
• Comprehensive Reporting: Generates detailed reports assessing vocabulary trends, profanity counts, and daily/weekly activity patterns.
• User-Defined Lexicon: Customize the list of swear words and phrases to fit personal or organizational standards by adding or overriding terms.
• Visual Trends: Easy-to-read charts visually represent the data, helping to identify patterns in language use over time.

How It Works

Agentic Swear Jar processes prompt histories from Claude Code and Codex CLI, focusing on small global histories instead of full session transcripts to maintain data integrity. Its core functionality includes:

• Efficient Parsing: The tool reads each JSONL file line-by-line, tokenizing prompts with a Unicode-aware regular expression and storing counts and statistics for analysis.
• Accuracy: Only whole token matches are considered, avoiding common substring matching mistakes.
• Customization: Users can edit the default dictionary of swear words to reflect their contexts or to include other datasets like those available in public repositories.

Usage Example

To analyze your coding history, run the following command:

./swear-stats --open

This command opens a report summarizing your coding sessions, including metrics like total matches and intensity counts.

For those looking to generate machine-readable data along with the HTML report, execute:

./swear-stats --json report.json --output report.html

Privacy and Control

Agentic Swear Jar prioritizes user privacy by processing plaintext histories while ensuring that the generated reports do not contain any raw prompts or session identifiers. When using the software, the information gathered is only for personal use and can be shared selectively as desired.

Conclusion

Designed with simplicity and effectiveness in mind, Agentic Swear Jar redefines how developers can approach their coding experiences. By providing insights into how language expresses the challenges of programming, it serves as both a humorous reflection and a useful tool for coders seeking to improve their craft.