Introducing AI Code Insights

Josh Beitler

Product manager

As adoption of AI coding tools matures, the key question for engineering leaders is no longer whether developers are using AI—it’s whether AI is being used effectively.

Leaders face critical questions that are difficult or impossible to answer with existing tooling: How are developers actually using these tools, and what are they spending on tokens? Are teams providing AI with sufficient context and clarity to generate useful output? Are codebases and systems structured in ways that AI can effectively navigate?

Today, we’re announcing AI Code Insights—a new set of capabilities from DX that give organizations real-time visibility into how AI tools are being used and where they’re falling short.

Meet AI Code Insights

AI Code Insights provides real-time visibility into AI-generated code across your organization—down to specific commits—along with deep insight into how developers are using AI tools day to day.

AI-generated code attribution. See exactly which commits and PRs contain AI-generated code. The new AI Code Overview report lets you explore critical SDLC lifecycle metrics bucketed by the percentage of AI in each PR—so you can start to understand how AI-generated code moves through your development process differently than human-written code.

Agent Experience insights. Surface insights directly from AI agents about where they’re hitting bottlenecks during a session—missing context, ambiguous instructions, structural barriers in the codebase. This is a fundamentally new signal: rather than inferring effectiveness from output, you’re hearing from the agent itself. The new Agent Experience report provides an overall agent effectiveness score across your organization, filterable by team, with a detailed per-session view of the score and its components.

AI dollar impact. Tie it all together with the AI dollar impact report, which translates your AI tool investment into an estimated net dollar figure. Rather than debating whether AI is “worth it” in the abstract, leaders can see a concrete financial picture of what their AI investment is producing—making board-level reporting and budget conversations significantly easier.

How it works

AI Code Insights is powered by a lightweight, self-hosted CLI daemon that runs on developer machines across macOS, Windows, and Linux. It supports all major AI coding tools—from IDE-based assistants like Cursor and Windsurf to CLI-based agents like Claude Code and Codex—using a layered approach to maximize accuracy across your AI toolchain.

For Claude Code, GitHub Copilot, and Cursor, the daemon plugs directly into native tool hooks—providing the most accurate data possible, including full session data such as token usage, individual conversations and prompts, and per-session Agent Experience insights. Native hook support for additional tools is on the roadmap.

For all other AI coding tools, the daemon falls back to a combination of local heuristics including real-time file change monitoring, giving you accurate, PR-level AI code attribution regardless of which tools your developers use.

The daemon is enterprise-ready. It can be deployed and managed through MDM solutions, making it straightforward to roll out across hundreds or thousands of developer machines without requiring individual setup.

AI Code Insights is designed with security at its core. No source code ever leaves the developer’s machine. The daemon performs all code analysis locally and transmits only aggregated metrics and metadata to DX—so organizations get full visibility into AI-generated code without exposing their codebase. Per-session data enrichment and Agent Experience insights are optional, configurable features.

Availability

AI Code Insights is generally available today for all DX Data Cloud customers. To learn more or get started, reach out to your account representative or request a demo.

Last Updated

April 9, 2026