Zozo

Every software project has a pile of small, clearly written tasks that wait for months because nobody has a free afternoon. AI coding agents can write that code today, but a person still has to start each session and watch it to the end. Zozo hands the pile to an agent of your choice and lets it work through the tasks unattended. Each finished task comes back as a pull request, a proposed change that a developer reads before it goes into the code.

Work starts with a conversation: you describe a feature, and Zozo asks one question at a time until it has a plan of linked issues. Once you confirm, it publishes them in order, and a task starts only after the ones it depends on are closed.

Each attempt runs on a separate branch, where the agent changes the code and runs your build and tests before committing. When the agent reaches an important decision nobody has written down, it pauses to ask, and the answer can come from a phone. Decisions end up in the repository as notes that later tasks can read.

Projects connect to GitHub, or to Bitbucket Cloud paired with Jira Cloud. The hosted version gives each customer a dedicated virtual machine in Germany or Finland. The site says code, git history and credentials stay on that machine. New hosted accounts are let in a few per day. A local version starts with one npx command on a Mac or Linux computer.