What It Does — Cost Estimation for AI Coding Projects
vibe-budget is a new CLI tool that solves a specific, painful problem for developers using AI assistants like Claude Code: unpredictable token costs. You describe your project in plain English, and it analyzes the tasks involved, estimates total token usage, and fetches real-time prices across 85+ models via OpenRouter.
For example, running:
vibe-budget plan ecommerce with stripe oauth and supabase
Might detect 4 distinct development tasks, estimate ~497k tokens, and then show you a cost comparison between the cheapest, best value, and premium model options.
Its scan command is particularly powerful for Claude Code workflows. You can point it at an existing codebase, and it will estimate how many tokens it would cost for an AI to refactor or extend that code.
Setup — Install in Seconds
Installation is straightforward via npm:
npm install -g vibe-budget
No API key is required to start estimating. The tool fetches live pricing from OpenRouter with a 1-hour cache as a fallback.
When To Use It — Before Every Major Claude Code Session
Integrate vibe-budget into your planning phase:
- Project Scoping: Before diving into a new feature with Claude Code, run
vibe-budget plan. Use the output to decide if you should break the work into smaller, more affordable chunks. - Refactor Planning: Use
vibe-budget scan ./your-projectbefore asking Claude to "refactor this entire module." The estimate tells you if a piecemeal approach would be more cost-effective. - Model Selection: The side-by-side price comparison helps you decide if using Claude 3.5 Sonnet is justified for a task, or if a less expensive model like Haiku would be sufficient, saving your Anthropic credits for more complex problems.
This tool turns cost from a surprise at the end of a session into a known variable you can optimize for at the start.
