Open-source optimAIzr targets coding-agent usage costs
A senior software engineer said on Reddit that they built optimAIzr with Claude Code while seeking to reduce repeated-context costs in coding-agent usage. The creator says the free, open-source tool runs locally without accounts or prompt uploads, analyzes usage across Claude Code, Codex, and opencode, and can help verify savings. Their personal transcript analysis attributed 56% of API-equivalent costs to rereading context, with cache expiry accounting for $98.74 over 27 days, but the report provides no independent verification.
The release gives Claude Code, Codex, and opencode users a local, open-source tool for inspecting usage and checking whether claimed savings materialize.
I’m a senior SWE using Claude Code daily. I analyzed my transcripts and found that 56% of API-equivalent costs went to rereading context. Cache expiry alone cost me $98.74 in 27 days.
So I built optimAIzr with Claude Code, using it to implement features, write tests, and iterate on the optimization logic as a development speed boost.
It analyzes usage, finds expensive patterns, and helps verify savings across Claude Code, Codex, and opencode.
Free, open source, runs locally. No accounts or prompt uploads.
npm i -g optimaizr git: https://github.com/blendbunjaku/optimaizr