New Relic ties production telemetry to GitHub Copilot and MCP Server

The problem: faster code, more incidents

New Relic argues that AI coding assistants such as GitHub Copilot have sharply increased developer velocity, but that shipping AI-generated software quickly creates new operational risks. It points to its own 2026 State of AI Coding report: 94% of engineering leaders rate AI-generated code as higher quality during review, yet 78% report an increase in production incidents after deployment.

The new integrations are meant to close that gap by bringing continuous telemetry into the developer workflow. That covers local coding in the integrated development environment (IDE), automated pull request (PR) verification and incident response, so that engineers and AI agents work from the same data.

Change Tracking and Autopilot: from PR to incident resolution

New Relic Change Tracking for GitHub Copilot builds a closed loop between production data and fixes. According to New Relic, it works in three steps:

• Before merge, it reads the PR diff and posts a monitoring plan as a PR comment.
• After every deploy, it checks performance at six points (10 minutes, 30 minutes, 1 hour, 2 hours, 24 hours and 72 hours) and returns a verdict: Good, Inconclusive, or Regression detected.
• If a regression is found, a GitHub Issue is logged automatically. GitHub Copilot reads the issue and generates a fix PR, and Change Tracking restarts the verification cycle once the fix is reviewed.

New Relic says enhancements are on the way, including streamlined integration onboarding, custom scheduling for monitoring runs, and more.

New Relic Autopilot for GitHub MCP Server targets incident response. The company notes that when an incident fires, the relevant code change sits in GitHub while production signals sit in New Relic, and correlating them by hand slows resolution. The integration feeds GitHub releases, PRs and commit history into Autopilot, shows development data next to performance metrics in one interface, and correlates repository events with telemetry anomalies to pinpoint the PR or line of code behind a regression. New Relic says this reduces mean time to resolution (MTTR).

Preflight: watching AI usage and cost locally

New Relic Preflight for GitHub Copilot moves observability into the developer's IDE and command-line interface (CLI). New Relic describes a lack of transparency into local AI usage and unmonitored model spend as hidden performance debt and a budget risk.

According to the company, Preflight tracks token consumption, execution latency, real-time spend and cost forecasts while code is being written. It normalizes telemetry across different LLMs and tools into one view for side-by-side comparison of performance, latency and cost. It also connects local data to New Relic dashboards for team-wide alerting, budget forecasting and long-term analytics.

What it means for development teams

For teams already using GitHub Copilot, the announcement points to a workflow where monitoring starts at the pull request rather than after an incident, and where regressions can be routed back to Copilot for a proposed fix. Teams concerned about AI model costs get visibility into spend before code is pushed.

New Relic will be at booth #114 at GitHub Universe in San Francisco on October 28-29. Setup instructions for Change Tracking are available at docs.newrelic.com, and Preflight for GitHub Copilot can be tried now, according to the company. The source does not give pricing or availability details beyond that.