Repo Tokens: GitHub Action Adds Token Count Badge for LLM Context Window Awareness
Repo Tokens is a GitHub Action that calculates the token count of your codebase and adds a visual badge to your README indicating how much of a large language model's context window your repository occupies. The goal is to make token size a visible metric, similar to bundle size badges for JavaScript libraries, encouraging leaner codebases that are more compatible with AI coding agents.
How It Works
The action uses tiktoken for tokenization. It runs approximately 60 lines of inline Python and takes about 10 seconds to execute. It's implemented as a composite action.
Configuration and Output
The context window size is configurable, defaulting to 200,000 tokens (matching the size of Claude models). The badge color reflects the percentage of the configured context window your codebase fills:
- Green: Under 30%
- Yellow: 50-70%
- Red: 70%+
The action updates the README file but does not automatically commit the changes, allowing your workflow to control the git strategy.
This type of tool is useful for developers working with AI assistants, as keeping a codebase within an LLM's context window allows the agent to maintain full awareness of the project structure and dependencies during coding tasks.
📖 Read the full source: HN AI Agents
👀 See Also
Google Releases Sashiko: AI Code Review Agent for Linux Kernel Patches
Google engineers have open-sourced Sashiko, an agentic AI code review system designed for the Linux kernel. It found 53% of bugs in an unfiltered set of 1,000 recent upstream issues that were missed by human reviewers.
ModelFitAI: Deploy AI Agents Without VPS Setup, Built with Claude Code
ModelFitAI is a platform that lets developers deploy AI agents directly on its infrastructure, eliminating VPS setup, Docker configuration, and SSH sessions. The entire platform was built using Claude Code by a solo founder.
Total Recall: Local Knowledge Graph for Claude Code Conversation History
Total Recall is an open-source system that ingests Claude Code's JSONL conversation transcripts into a SQLite database with full-text search and vector embeddings, making conversation history searchable across sessions. It retrieves actual conversation excerpts with DAG-aware context and includes a ChatGPT importer.
Claude Code Routines: Schedule Agent Tasks Like Cron with Reasoning
Claude Code Routines let you run agent tasks on a schedule without keeping a session open. A Reddit user shares real examples: nightly commit review, weekly dependency check, daily error log analysis — with AI reasoning instead of raw script output.