Noddit

Noddit is an open-source tool from WeCarbon that gives AI agents one place to log their work and ask questions, and gives the human owner one screen to watch and answer them. Agents write tasks and questions as markdown files in a project's decisions/ folder through the nod CLI, and the owner answers from a live terminal bench or a desktop app. It is released under the MIT license, and the latest listed release is v0.1.1.

What It Is

Noddit is a coordination layer between running agents and their owner, not a planning board. Everything is a single kind of item, a node: something an agent is doing. When an agent needs the owner first, the node becomes a decision with options, a default, and a note on what happens if nobody answers. The owner answers with a keypress, the agent is notified, reads the answer with nod read, and carries on.

How the Workflow Runs

An agent creates nodes with nod add, splits work into child nodes with --after dependencies, reports progress with nod progress (including parsing pytest output from a pipe), and asks questions with nod ask. The bench (nod or nod bench) offers Available, Active, Decide, All and Waiting views, with progress bars and staleness coloring: grey after 30 minutes without an update, red after an hour. nod monitor streams notifications to agents so they do not have to poll.

Storage and Safety Model

Each node is its own markdown file with YAML front matter, versioned in git. Ids are reserved with O_EXCL files so many agents can write at once without collisions. Approval gates record the sha256 of an attached draft, and nod approved --check succeeds only if the owner picked the approving option for exactly that content. External sources such as GitHub issues or a support inbox sync references only (ids and titles), not bodies or credentials.

Setup Path

Install with npm i -g @wecarbon/noddit (Node 22.18+), then run nod init in a project. A Claude Code plugin injects usage rules at session start and starts the monitor. Other agents can follow the rules in AGENTS.md, since they are plain CLI calls. A desktop app for macOS, Windows and Linux and a Python SDK with a pytest plugin for reporting progress from containers are also provided. The UI is available in English, Simplified and Traditional Chinese, French and Arabic.