Continue Launches 'Agents' Beta for Running AI in the Background

Continue has launched Agents in beta status, a feature that supports AI workflow orchestration running continuously throughout the development lifecycle. Unlike existing AI coding tools that have been locked into a request-response paradigm, Agents reflects the actual development environment where tasks like testing, building, and deployment operate asynchronously in the background.

Differences Between IDE Agents and Cloud Agents

IDE Agents operate synchronously and interactively on the local machine, optimized for Copilot-style assistance. In contrast, Cloud Agents run asynchronously and autonomously in the cloud or a separate process. They are suited for long-running tasks such as refactoring, documentation updates, and security patches without blocking the terminal.

Use Cases and Getting Started Strategy

Agents are used for handling minor bugs during architecture work, generating pattern-based boilerplate code, investigating issues using full context, and automating security scans. Users connect their GitHub account at hub.continue.dev/agents and enter a prompt, after which Continue creates a PR. The recommended strategy is to start with small bug fixes that have clear solutions, build up a track record of success, and then gradually expand to more complex tasks.

Workflow Changes and Feedback

Since Agents run for extended periods, clearly conveying important details upfront is key to success. As the volume of PRs generated increases, teams need to adjust their code review habits and discuss where existing practices break down due to AI. Continue provides full control over models, rules, and tools, and is collecting community feedback through GitHub Discussions and Discord.