Codag

Founded by Michael Zixi Zhou

Reduce coding agent token costs by compressing verbose outputs and logs

Pricing model

Free Tier

Rating

★ People love it

Time to value

Quick Setup (< 1 hour)

You can use Codag to lower the cost of running AI coding assistants like Claude Code and Codex. It runs locally as a proxy and filters large command outputs, test runs, build logs, and search results before sending them to the model. Passing test suites and repetitive noise are trimmed down to only the relevant failures and matches. If the agent needs omitted details, it retrieves them directly from local storage. Your source code, diffs, and prompts remain completely unchanged. It also tracks saved tokens across developer sessions.

What Codag does

Compress command-line test outputsFilter terminal build logsInspect token savings with terminal analytics commandsRetrieve omitted output lines on demandRun local proxies for coding agentsSet up coding agent cost controlsAttaches to Claude Code and Codex using a single setup commandFilters test runs, build logs, and search results to show only critical linesLeaves source code, diffs, and configuration files completely untouchedStores full uncompressed output locally and encrypted for on-demand retrievalRoutes model traffic directly through a local proxy to your providerFails open during network issues or quota limits to prevent breaking active sessionsTracks avoided tokens and estimated dollar savings per developer session