Cloudflare introduces Issues for Workers to automate error monitoring and agent handoff

Streamlining the agent workflow

Cloudflare has introduced Issues, a built-in error monitoring feature for Cloudflare Workers now in open beta. This feature addresses the manual gap between detecting production failures and investigating them by grouping repeated exceptions, 5xx responses, and error logs into single issues, thereby eliminating the need for agents to reconstruct failure context from raw telemetry.

Automated detection and context

Embedded directly into the Workers runtime, Issues requires no SDK installation or application wrapper. With a single configuration line, it captures uncaught exceptions, failed invocations, HTTP 5xx responses, and logs containing stack traces. It also flags runaway alarm conditions and excessive logging in loops. The system groups related errors to display their first appearance, frequency, and trends. Users can enrich this data with application-specific identifiers (e.g., user or session IDs) using the built-in OpenTelemetry API, allowing agents to identify if failures are concentrated in specific accounts or sessions.

Automated detection and context

When an issue crosses a defined occurrence threshold or returns after a quiet period, Automations can send the complete diagnostic context—including stack traces, logs, traces, and Worker versions—directly to configured coding agents. Supported integrations include:

• Claude Code (via routine ID and token)
• Cursor (via automation webhook URL)
• Devin (via API token and organization ID)
• Generic webhooks and chat/incident management tools

Agents can further investigate by connecting to Cloudflare MCP to query related logs and traces, propose code changes, and open pull requests for human review.

Real-world application

Cloudflare tested Issues on Cloudflare Workflows, a platform built on Workers. Within a day of activation, the team identified two hidden bugs: a control plane migration stuck in a retry loop due to a SQLite foreign key error, and a deletion process that failed to complete due to Workers subrequest limits. The automation setup sent these issues to Cloudflare OS, which analyzed the code and proposed fixes for both problems.