About todox — working memory for developers and their agents
todox is not a checklist. Each task carries a log — the decisions behind it, the approaches that failed, the questions still open, and the note the last session left behind. A fresh agent reads it with one get_context call and continues without asking you anything.
What it is
- A project per repository, with the tasks you are working on and the log entries that explain what happened to them.
- A live log per task: doing, blocked, done. Each change is a row, not an edit, so the history is the source of truth.
- A daily and weekly activity report written for whoever is starting their day — the developer or the agent they are about to hand off to.
- An MCP server (/api/mcp) that exposes the same surface to coding agents. The hosted endpoint and the stdio process share one tool list.
What it is not
- A kanban board. There is no swimlane, no sprint, no Gantt.
- A team chat. Comments live next to the change they describe.
- A wiki. Notes are short and dated.
- A replacement for your agent's built-in memory. Claude Code's auto memory is per repository, per machine and shared with nobody, by its own documentation; todox is the log that crosses those lines — machines, agents and people.
Built by
todox is built and maintained by Furkan Beydemir. The repository is open source under the MIT license, and contributions are welcome on GitHub.