Token Limit helps you monitor how many tokens your AI context files consume. Set token budgets for your prompts, documentation, and configs, then get alerts when limits are exceeded.

Keep your AI costs predictable and avoid hitting context window limits that break your applications.

AI context files are becoming a standard part of modern development workflows. Projects now commonly include .context/, CLAUDE.md, .clinerules, .cursorrules, and other AI instruction files directly in their repositories.

As these files grow in size and complexity, it becomes crucial to monitor their token consumption to avoid unexpected API costs and context window limitations.

• Multi-model support for OpenAI GPT and Anthropic Claude
• CI integration to catch budget overruns in pull requests
• Flexible configuration for different AI use cases
• Real token costs instead of inaccurate file sizes
• Cost budgets in dollars and cents, not just tokens
• Up-to-date pricing from OpenRouter API instead of hardcoded values

1. Configure your token budgets in token-limit.config.ts, package.json, or other supported formats
2. Analyze files using official tokenizers for each AI model (tiktoken, Anthropic)
3. Report which files exceed limits with detailed breakdowns
4. Prevent costly overruns by failing CI builds when budgets are exceeded

1. Install Token Limit:

npm install --save-dev token-limit

2. Create a configuration file (e.g., token-limit.config.ts or .token-limit.json):

// token-limit.config.ts

import { defineConfig } from 'token-limit'

export default defineConfig([
  {
    name: 'AI Context',
    path: '.context/**/*.md',
    limit: '100k',
    model: 'gpt-4',
  },
  {
    name: 'Documentation',
    path: ['docs/**/*.md', 'docs/**/*.txt'],
    limit: '$0.05',
    model: 'claude-sonnet-4',
  },
])

3. Add a script to your package.json:

{
  "scripts": {
    "token-limit": "token-limit"
  }
}

4. Run the analysis:

npm run token-limit

You can also run Token Limit directly from the command line:

# Check specific files
npx token-limit README.md docs/guide.md

# Set limits and models
npx token-limit --limit 10k --model gpt-4 docs/**/*.md

# Set cost limits
npx token-limit --limit '$0.25' --model gpt-4 expensive-prompts/**/*.md

# Name your check
npx token-limit --name "API Docs" --limit 50k api-docs/**/*.md

# Multiple examples
npx token-limit .context/**/*.md
npx token-limit --limit 1000 claude.md
npx token-limit --limit '5c' --model gpt-3.5-turbo quick-prompts/*.txt
npx token-limit --json --hide-passed

Token Limit supports multiple configuration formats to suit your project needs. You can define token limits, models, and file paths in a variety of ways:

• token-limit.config.{ts,js,mjs,cjs}
• .token-limit.{ts,js,mjs,cjs,json}
• .token-limit
• package.json (token-limit field)
• Command line arguments

OpenAI Models

• gpt-5
• gpt-4.1
• gpt-4.1-mini
• gpt-4.1-nano
• gpt-4o
• gpt-4o-mini
• gpt-4-turbo
• gpt-4
• gpt-3.5-turbo
• o3-mini
• o1

Anthropic Models

• claude-opus-4.6
• claude-opus-4.5
• claude-sonnet-4.5
• claude-haiku-4.5
• claude-opus-4.1
• claude-opus-4
• claude-sonnet-4
• claude-3.7-sonnet
• claude-3.5-sonnet
• claude-3.5-haiku

Token Limits

• Numbers: 1000, 50000
• Human-readable: "10k", "1.5M", "500K"

Cost Limits

• Dollar amounts: "$0.05", "$1.50"
• Cents: "5c", "10c"
• Plain numbers: 0.05, 1.5 (interpreted as dollars)

Add Token Limit to your CI pipeline:

# .github/workflows/token-limit.yml

name: Token Limit
on: [push, pull_request]

jobs:
  token-limit:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: '18'
      - run: npx token-limit

Unlike traditional bundle size limits, token limits directly impact:

• API Costs: More tokens = higher bills (GPT-4 costs $0.03 per 1K tokens)
• Response Quality: Exceeding context windows truncates input (GPT-4: 128K limit)
• Performance: Larger contexts mean slower API responses
• Reliability: Context overflow can cause API errors

Token Limit helps you catch these issues before they reach production.

See Contributing Guide.

MIT © Azat S.