New Relic brings production performance data into GitHub Copilot

Closing the feedback loop

Software engineering teams are increasingly using AI tools like GitHub Copilot to accelerate the pace of development. While these tools help teams produce code more rapidly, the speed of creation can sometimes outpace the ability to verify quality. New Relic is addressing this by integrating production data directly into the developer's primary workspace.

According to New Relic's 2026 State of AI Coding, there is a significant gap between human review and actual performance. While 94% of leaders rated AI-generated code as high quality during the review phase, 78% of those same leaders reported an increase in incidents once the code was actually deployed. This suggests that code which looks correct to a human or an AI can still cause issues in a live environment.

The goal of these new integrations is to eliminate the need for developers to switch between different tools to find out why an application is failing. By bringing telemetry—the data regarding a system's health and performance—into the coding workflow, engineers can see the consequences of their work in real-time. This helps prevent the cycle of prolonged triage where teams spend hours searching for the source of a performance anomaly.

How Change Tracking works

New Relic Change Tracking for GitHub Copilot functions by connecting production telemetry with the deployment workflow. It provides a continuous stream of data that allows engineers to see how newly merged code affects downstream services. This visibility is designed to minimize manual triage and keep developers focused on building features rather than hunting for bugs.

The system operates by running several types of checks against a performance baseline. It can perform a pre-merge check to see if code is ready, a canary gate to test a small portion of traffic after deployment, or regression detection to find issues after a full rollout. All of these checks are performed against real-time performance data rather than static tests.

When the system identifies a regression, it automates the notification and remediation process. It automatically routes the issue to GitHub Issues and Copilot for the necessary fixes and sends updates to Slack. This ensures that the team is immediately aware of the problem and has a clear path to resolve it without manual intervention.

Context for incident response

Beyond tracking changes, New Relic is also introducing tools to help with incident response using the Model Context Protocol (MCP). This is an open standard that allows different AI models and tools to share context. The New Relic Autopilot integration for GitHub MCP Server connects an SRE (Site Reliability Engineering) agent to a company's specific codebase.

When an alert occurs, the Autopilot agent can draw on the context of the entire codebase to triage the problem. It identifies the root cause and recommends specific fixes to the developers. This provides clearer guidance during high-pressure situations, allowing engineers to resolve incidents faster while staying within their existing tools. It moves the intelligence from a closed dashboard into the active workspace.

The integration aims to give AI agents a shared understanding of how the software behaves in production. Camden Swita, Head of AI at New Relic, said engineers and agents need a shared understanding of software behavior. By providing this context, the tools help turn faster coding into more reliable software delivery.

Monitoring the AI itself

New Relic is also providing tools to help teams understand how they are using AI coding agents in the first place. The New Relic Preflight integration for GitHub Copilot is an open-source capability designed to give developers visibility into the coding process itself. It tracks how efficiently agents are being used and identifies any errors that occur during execution.

Developers can view local telemetry directly in their integrated development environment (IDE) or command line interface (CLI). This includes real-time data on model efficiency, latency, and token consumption. Token consumption is a key metric for teams looking to manage the costs associated with using large language models for development.

With this unified view, engineering teams can make informed decisions about which models to use. They can optimize their costs, set up proactive alerts for performance issues, and evaluate the accuracy of suggestions across different models. This ensures that the adoption of AI tools is both efficient and economically sustainable for the organization.