Deadweight Tool Reveals Hidden Claude Code Token Costs

A new open-source tool called Deadweight is helping developers understand exactly how much their Claude Code workspace configuration costs before writing a single line of code. Released on GitHub on November 19, 2026, the analyzer scans your entire context window setup and calculates the token overhead that gets sent with every request, exposing hidden costs that many users never realized they were paying.
How Deadweight Works
The command-line tool performs a static analysis of your Claude Code environment, examining every file in your context window, custom instructions, MCP server integrations, and system prompts. It then outputs a detailed breakdown showing exactly how many tokens are consumed by each component before Claude even sees your actual coding request. The analyzer runs locally and requires no API calls, making it safe for sensitive codebases.
According to the project documentation, Deadweight categorizes overhead into four buckets: workspace files automatically included in context, custom instructions and rules you've configured, MCP tool definitions and schemas, and system-level prompts injected by extensions. Each category gets a token count and percentage breakdown, allowing developers to identify the biggest contributors to their per-request costs.
Release Date and Availability
Deadweight was officially released on November 19, 2026, through the msertdev GitHub repository. The tool is distributed as an MIT-licensed open-source project with installation support for macOS, Linux, and Windows environments. Developers can install it via npm or download pre-compiled binaries from the releases page.
Real-World Cost Findings
Early adopters on Hacker News are reporting significant discoveries. One user found 4,200 tokens of overhead from a forgotten .cursorrules file that was being sent with every request, costing roughly $0.012 per interaction at standard Claude Code pricing rates. Another developer discovered that their MCP server definitions alone consumed 1,800 tokens per request, primarily from verbose JSON schemas that could be condensed without losing functionality.
The tool has already identified common cost culprits: workspace-wide ESLint and Prettier configuration files that rarely need to be in context, overly detailed custom instructions that repeat information already in system prompts, and MCP integrations that load full API documentation even when only using basic functions. Some users are reporting 30-40% reductions in their monthly Claude Code bills after using Deadweight to audit and trim their configurations.
Integration with Existing Workflows
Deadweight outputs its analysis in both human-readable and JSON formats, making it easy to integrate into CI/CD pipelines or pre-commit hooks. Several developers are using it to enforce context budget limits in team environments, failing builds when overhead exceeds a defined threshold. The tool also supports diff mode, comparing two configurations to show how changes impact token usage before deploying them to production environments.
What This Means
As AI coding assistants become core development infrastructure, understanding the true cost of your configuration is no longer optional. Deadweight provides the visibility needed to optimize Claude Code setups without sacrificing functionality, particularly for teams running hundreds or thousands of requests daily. The tool's release signals a broader trend toward cost transparency in AI tooling, where developers need granular control over what they're paying for with each interaction. For organizations scaling Claude Code usage, an audit with Deadweight should become part of the standard setup process.
on Emergent today





