Continue Unveils 'Continue Agents,' AI Agents Tailored to Team Standards

Continue has unveiled Continue Agents, which go beyond general-purpose chat assistants to reflect a team's custom context and rules. These agents focus not on text generation but on executing complex workflows, and are composed of a combination of Prompt (mission), Tools & MCPs (capabilities), and Rules (team standards).

Agents can run in the CLI, Mission Control, and automation environments, guaranteeing consistent output regardless of who or what is executing them. Cloud Agent is triggered on remote infrastructure by tasks, schedules, or external events, producing reviewable results across shared engineering systems.

Key Features and Composition

• Consistency: Whether it's a junior developer or CI automation, the agent operates by the same standard according to the team's format and rules.
• Shareability: Once defined on the Hub, the entire team can invoke it without needing to re-explain team standards.
• Governance: Rules and guardrails are built in, making automation safe and auditable.

Execution Modes and Use Cases

Agents are divided into Interactive Mode (terminal conversational) and Headless Mode (automation). Interactive Mode is suited for complex debugging or configuration testing, while Headless Mode handles repetitive tasks—such as scheduled security scans, PR reviews, and documentation updates—without manual intervention.

For example, the Snyk Security Scan agent discovers vulnerabilities, creates issues in Linear, and proposes fix PRs. GitHub Issue Manager triages open issues and automatically generates release notes. These agents can be installed and run with the command npm i -g @continuedev/cli.