Overview
Dobi makes the work you repeat with your agent better and cheaper, and keeps the evidence.
You work in Claude Code or Codex as you already do. When a piece of work is worth repeating, ask your agent to use Dobi. It saves the work as a task, with example inputs and what a good result looks like. Each time the agent runs the task, Dobi records the run: its output, how it was judged, and the tokens it used. You can then compare versions of the task and keep the one that does the job for less.
What runs where
Section titled “What runs where”- Your agent does all the work. It runs on your own subscription or API keys. Dobi’s service never makes a model call.
- The plugin runs on your machine. It records runs into files only your user can read, under
~/.local/share/dobi. - The app shows your tasks, runs and results, once you connect a device. It stores and displays evidence, and never starts work.
Private vs Synced says exactly what leaves your machine. Dobi is a free preview, so for your company’s or your customers’ data, use Private for now.
