Google Jules CLI: Run async agents from your terminal (www.npmjs.com)

Google has released a Jules CLI that brings its asynchronous coding agent into the terminal, letting developers spin up, manage and retrieve results from remote agent sessions directly from their shell. The tool includes a TUI entrypoint (jules), authentication (login/logout with your Google account), autocompletion, and simple commands for remote session lifecycle: jules remote new to create sessions tied to a repo, jules remote list (--session/--repo) to inspect them, and jules remote pull --session <id> to fetch outputs. Examples show batch workflows (creating one session per TODO line), targeted workflows (create a session from the first GitHub issue assigned to you), and composition with other CLIs like gh and Gemini to triage or pre-process issues before sending them to Jules. For the AI/ML community this matters because it operationalizes async agent workflows into everyday dev toolchains: background code-writing, test generation, issue triage, and repo-scoped automation become scriptable and reproducible. Technical implications include smoother integration with CI/DevOps pipelines, ability to parallelize agent tasks across sessions, and easier orchestration via shell pipelines. It also raises practical considerations around authentication, repo access permissions, remote compute costs and reproducibility of agent runs—important for secure, auditable deployment of agent-driven automation.