New Relic has announced the development of New Relic AI Coding Observability, an open-source new feature designed to bring observability to the use of AI programming assistants. The proposal addresses a growing issue within engineering teams: increasingly, code is originating from tools like Claude Code, Cursor, GitHub Copilot, Windsurf, or Amazon Q, yet many companies lack real visibility into how these tools are used, how much they cost, their impact on productivity, or the risks they introduce into the development cycle.
The company, specialized in observability, presents this new capability as a way to extend control beyond production. Until now, enterprise observability has focused on deployed applications, infrastructure, APIs, traces, logs, metrics, and user experience. With AI-assisted programming, a significant portion of risk appears earlier in the process: during the phase when code is generated, accepted, modified, and committed to corporate repositories.
Programming AI Becomes a Blind Spot
The use of coding assistants is growing rapidly. New Relic cites a Gartner forecast indicating that by 2028, 90% of enterprise software engineers will use AI coding assistants. This aligns with a visible trend: companies are no longer questioning whether their developers will use AI but are instead focused on how to monitor that usage without hindering productivity.
The challenge is that few organizations work with a single tool. One team might use GitHub Copilot in their IDE, another Cursor for prototyping, another Claude Code for refactoring, Amazon Q in cloud environments, and Windsurf for specific tasks. This fragmentation complicates governance, as each tool has its telemetry, usage model, costs, and risks.
New Relic AI Coding Observability aims to normalize this information into a common, provider-agnostic view. The company seeks to correlate the usage of programming assistants with existing data in the production infrastructure. The simple but powerful idea is to understand whether code generated by AI truly improves productivity, shortens delivery times, reduces errors, lowers development costs, or introduces patterns that might later affect production.
| Area | What New Relic AI Coding Observability Proposes | Why It Matters |
|---|---|---|
| Usage visibility | Monitor programming assistant actions | Avoid reliance on perceptions or internal surveys |
| Costs | Track expenses, budgets, and thresholds | Reduce opaque invoices related to AI usage |
| Productivity | Measure actual impact on development | Replace anecdotal stories with data |
| Security and compliance | Local-only mode and zero outbound | Assist with privacy, data sovereignty, and regulations |
| Transparency | Open source and readable code | Enable audits by engineering and security teams |
| Portability | Support for OpenTelemetry and MCP | Reduce dependence on a single provider |
| Coverage | Claude Code, Cursor, GitHub Copilot, Windsurf, Amazon Q | Reflects the fragmented reality of teams |
Cost, Productivity, and Compliance in the Same Dashboard
New Relic’s proposal addresses three concerns common to engineering management: cost, productivity, and control. While AI assistants can speed up tasks, they are also becoming a new expense line that’s hard to attribute. Licenses, consumption, team usage, duplicated tools, and varying models can generate opaque bills.
According to the company, this new feature will allow monitoring AI assistant spending, forecasting consumption against budgets, and launching alerts before thresholds are exceeded. This dimension is crucial because development AI should not be seen solely as a technical improvement but also as an economic resource requiring discipline, similar to cloud, observability, or data platforms.
Regarding productivity, New Relic aims to move from enthusiasm to measurement. Many companies have internal stories of developers delivering earlier, reducing repetitive tasks, or speeding up testing thanks to AI. However, there is often no reliable way to verify if these improvements are widespread, sustainable, and compatible with software quality. Applying observability to AI-assisted coding could reveal inefficiencies, unproductive usage, excessive dependencies, or patterns that seem to accelerate development but generate technical debt later.
The third pillar is compliance. New Relic announces a local-only or zero outbound mode that runs queries inside the user’s private network. This feature could be important for regulated companies, organizations with sensitive intellectual property, or teams that don’t want to expose code snippets, prompts, metadata, or telemetry to external services. The company also emphasizes that the code will be open source or source-available, enabling teams to verify how their data is managed.
OpenTelemetry and MCP as a Neutral Foundation
One of the most interesting technical decisions is the native support for OpenTelemetry and the Model Context Protocol, MCP. OpenTelemetry has become the de facto standard for collecting and exporting telemetry data in modern applications. MCP, on the other hand, is gaining traction as a protocol to connect AI assistants with tools, context, and external data sources.
By supporting both, New Relic aims to prevent AI Coding Observability from being locked into a single ecosystem. The goal is for organizations to port their telemetry and AI workflows across clouds, languages, and model providers. In a still rapidly evolving market, this neutrality could hold value: no one knows which assistant will dominate in two years, or whether companies will end up using a mix of proprietary, open-source, local, and cloud models.
Initial coverage will include Claude Code, Cursor, GitHub Copilot, Windsurf, and Amazon Q. The future addition or removal of tools will be significant because the developer assistant market is very dynamic. The success of an observability solution at this layer will depend on its ability to adapt to a set of tools that change almost every quarter.
Observability Moves to the Left
New Relic’s announcement aligns with a broader trend: control tools are shifting further earlier in the software lifecycle. Security has been talking about shift-left for years. Now, observability is following the same path. It’s no longer enough to know a fault occurred in production; it’s also vital to understand how the code was generated, which assistant was involved, what decision was accepted, and what costs were incurred to reach that point.
This change does not replace traditional engineering practices. Code reviews, testing, SAST, DAST, dependency analysis, production observability, and incident response remain essential. But AI adds a new layer: the behavior of the assistant as part of the development system.
New Relic AI Coding Observability will be available as an open-source feature on June 23, without additional cost for the feature itself, though standard New Relic ingestion fees apply. The local-only mode will be available soon, according to the company.
For technical teams, the message is clear. AI-assisted programming can no longer be managed solely as an individual tool installed by each developer. When it begins to impact costs, security, productivity, and software quality, it requires its own observability. The next stage isn’t just about writing more code with AI but about understanding whether that code is genuinely improving the system in production.
Frequently Asked Questions
What is New Relic AI Coding Observability?
An open-source ongoing project to observe the use of AI programming assistants, measure their impact, and correlate activity with the software lifecycle.
Which code assistants will it support?
Support for Claude Code, Cursor, GitHub Copilot, Windsurf, and Amazon Q is announced.
When will it be available?
It will be available on June 23 as an open-source project, without additional cost, though standard ingestion fees apply.
Why is this important for regulated companies?
Because it includes a local-only or zero outbound mode to run queries within the user’s private network, which is relevant for privacy, data sovereignty, and compliance.
via: newrelic