Vercel Agent expands from code review to incident response
At its Ship AI event, Vercel launched Agent Investigations in Public Beta, extending Vercel Agent beyond AI-powered code reviews into production issue detection, root cause analysis, and remediation planning. The new capability pairs with the recently released anomaly alerts to automate parts of the incident response workflow natively on the Vercel platform.
Why manual incident response no longer scales
Modern web applications generate enormous volumes of logs and metrics from every deployment, user interaction, and API call. When something goes wrong, someone has to dig through that data to find the cause. As shipping cadence accelerates, those manual investigations become a recurring drain: poorly tuned alerting desensitizes teams to real problems, hours of context switching erode productivity, and post-incident documentation frequently gets deferred in favor of catching up on delayed work.
Vercel's approach is to give the agent broad visibility across build time and runtime, so it can perform the kind of analysis a senior engineer would — but in seconds instead of hours, and without requiring third-party tooling.
Automatic anomaly detection with zero setup
Anomaly alerts continuously watch for unusual activity in your application, including spikes in billable metrics like function duration or fast data transfer, as well as increases in 5xx errors. No configuration is required: any team on Observability Plus gets anomaly alerts out-of-the-box. When an anomaly is detected, Vercel Agent steps in to investigate either automatically or on demand.
Agent Investigations then processes multiple streams of data simultaneously to determine what changed and why. It examines:
- Correlation analysis: which other metrics changed around the same time, whether a deployment just completed, or if traffic patterns shifted
- Historical context: whether the issue has occurred before and what resolved it last time
- Dependency mapping: which upstream or downstream services might be affected, and whether this is a cascading failure
- Change attribution: what code changes or configuration updates preceded the anomaly
The agent generates a succinct summary of what went wrong and why, producing actionable steps tied to the specific root cause it identified. This approach also helps teams triage: not every alert requires immediate action. Investigations assesses severity and impact, and if the issue has already resolved itself, the agent flags that so teams can focus elsewhere.
Availability and getting started
Agent Investigations requires Observability Plus and runs on Vercel Agent credits. New Agent users can claim $100 in free credit toward both Investigations and Code Reviews. Teams can enable the agent to run automatically on all error alerts via the Agent tab in the Vercel dashboard, or trigger an investigation manually.
Anomaly alerts are enabled by default for Observability Plus subscribers, with email alerts sent to team owners by default and optionally to team members. Slack integration and webhooks are available for connecting to existing communication, ticketing, and incident response tools.
