Developers relying on Anthropic’s flat-rate Claude Code subscriptions can now see exactly how much value they are extracting from the service. A new open-source plugin, "what-would-it-cost," sits directly in the terminal interface to calculate real-time expenses based on API list prices. The tool breaks down costs by model, token type, and session duration, offering a stark contrast between the fixed subscription fee and the variable cost of pay-as-you-go usage.

Real-Time Cost Visibility

The plugin provides an always-on status line that updates after every model request, showing the current session cost in dollars. It visualizes the model mix using colored indicators for Opus, Fable, Sonnet, and Haiku, while also tracking daily, weekly, and billing cycle totals. For Max 20x subscribers, the tool explicitly calculates the multiplier, showing how many times over the subscription cost the user has already gone in the current billing cycle.

Comprehensive Pricing Breakdown

The calculator uses Anthropic’s October 2026 list prices, including specific rates for input, output, and cache operations across models like Fable 5.1 and Opus 5.5. It distinguishes between 5-minute and 1-hour cache writes, although live usage defaults to the 1-hour rate for simplicity. The tool reads local transcripts from ~/.claude/projects to price historical data, ensuring that prompt caching savings and subagent usage are accurately reflected in the final bill.

Privacy and Local Processing

Privacy remains a core feature of the implementation, with no network requests or telemetry sent to Anthropic or third parties. All calculations occur locally, storing only token counts, timestamps, and model identifiers rather than the actual message content. The plugin identifies the user’s specific plan tierβ€”Pro, Max 5x, or Max 20xβ€”by reading local configuration files, allowing it to correctly contextualize the subscription cost against the calculated API value.

Key Takeaways

  • The plugin provides immediate feedback on whether heavy usage justifies the subscription price versus API billing.
  • It supports detailed pricing for new models like Fable 5.1 and Opus 5.5, using October 2026 rate cards.
  • All data processing is local, with no message text stored or transmitted, preserving developer privacy.
  • The tool tracks cache savings and subagent costs, which are often obscured in standard subscription billing.

The Bottom Line

This tool is a necessary transparency layer for power users who need to know if they are getting a bargain or overpaying for their flat-rate AI access.