Beyond the Editor: How I'm Using Continue CLI to Automate Everything

AI coding tools are evolving beyond simple function autocompletion, with Continue CLI (cn) addressing the real complexity of building applications. Agents can now proactively handle repetitive, structural tasks—security layers, architecture, testing, CI/CD—that existing AI previously couldn't manage.

Safe Automation Based on Security and Permissions

Continue CLI ensures automation safety through a permission system. Users can allow specific command patterns like cn --allow Bash(git*), or require confirmation before executing risky commands like cn --ask Bash(rm*). These settings are stored in the ~/.continue/permissions.yaml file and shared across teams, and can be easily integrated into scripts and workflows via pipe connections. For example, the result of git diff can be fed as input to automatically create a GitHub issue for an error.

Asynchronous Agents and Continuous AI

The core concept is Continuous AI, where AI operates even when the user isn't at the keyboard. The author built a triage bot integrated with GitHub Actions, where a Continue review agent automatically classifies and labels new issues and writes an initial response when they're created. This bot handles 80% of initial issue responses, and is planned to expand into a fully asynchronous workflow (Level 3) going forward—including automatic documentation updates on code changes and overnight test case generation.

Adoption and Use

Developers can install the CLI with npm i -g @continuedev/cli and start with simple automation by integrating it with GitHub CLI (gh). The shift in thinking is moving from "how can AI help with coding" to "what can AI do when I'm not around." On the roadmap are plans for reducing the frequency of human intervention, a feedback-learning rule engine, and enterprise audit trail features.