New Relic MCP Server

The New Relic MCP Server connects AI assistants such as Claude Code, Codex, Claude Desktop, or VS Code to the New Relic observability platform. Anyone monitoring applications and infrastructure with New Relic can ask the assistant questions like "Which errors have occurred since the last deployment?", "Which alerts are open right now?", or "What is slowing down this transaction?" without switching between the editor and the web interface. The server is operated by New Relic itself. According to the documentation, the core features are generally available, while individual tools are marked as preview.

What is the New Relic MCP Server?

New Relic stores telemetry such as metrics, events, logs, and traces and exposes it through the query language NRQL. The MCP server bundles a larger number of tools with which an agent can explore and analyze this data. According to the documentation the address is regional: https://mcp.newrelic.com/mcp/ for the US region, https://mcp.eu.newrelic.com/mcp/ for the EU, and https://mcp.jp.newrelic.com/mcp/ for Japan. You have to choose the address that matches the region of your account.

Which tools are available?

The tools are grouped. Exploration covers listing accounts, dashboards, and metric names, searching entities, and showing related entities. For data access there is execute_nrql_query and, via OAuth only, natural_language_to_nrql_query, which translates a plain-language question into an NRQL query. In the alerting area you can retrieve conditions, policies, incidents, and open issues, plus synthetic monitors and their results. For troubleshooting, the reference names tools for error groups, error statistics, and change events as well as the OAuth-bound analyses analyze_deployment_impact, generate_alert_insights_report, and generate_user_impact_report. For performance questions there are analyses of logs, golden metrics, Kafka, threads, transactions, and garbage collection. According to the reference, three trace tools are still in preview and read-only.

What is it particularly useful for?

The value lies in production context during development. After a deployment, an assistant can check whether error rates or response times changed, match error groups against the code in the editor, and narrow down causes. Compared with generic database servers, this server knows the platform's concepts, such as entities, alert policies, and golden metrics, and returns answers that fit the monitoring environment directly. Translating questions into NRQL also helps teams that do not use the query language every day.

Setup

According to the documentation, signing in via OAuth is recommended. In Claude Code, claude mcp add --transport http newrelic https://mcp.newrelic.com/mcp/ is enough, followed by the browser flow. The documentation names minimum versions for Claude Code and Codex as well as setup paths for Claude Desktop via the mcp-remote proxy, for VS Code, and for Kiro CLI. An API key is a second route, but the setup page describes it only for the Antigravity CLI, and some tools are not available with it. With the HTTP header include-tags you can restrict the tools by categories such as discovery, alerting, and data-access.

Security and limits

According to the documentation the assistant acts on behalf of your user. Use an account with the lowest permissions possible and restrict the tools with include-tags if the assistant should only read. Retrieved logs and error data enter the AI client context and therefore reach the model provider; they can contain personal data or credentials. The preview tools may still change.

Costs

According to the documentation the server itself is usable for all accounts; the core tools are open to all data accounts, while the full tool set is available to certain consumption models. According to its pricing page the platform offers a permanently free entry tier as well as paid plans with usage-based billing and an enterprise tier. Which terms apply to you is shown on the official New Relic pricing page.

Who is the New Relic MCP Server for?

It is worthwhile for teams that already use New Relic in production and want to connect their development tools with production data, such as development, SRE, and DevOps teams. Anyone without New Relic data gains nothing from it: the server is not a standalone monitoring system but an access path to the platform.

A New Relic account in the matching region and an MCP client that supports remote servers, such as Claude Code, Codex, Claude Desktop, or VS Code.

With the New Relic MCP Server, an AI assistant queries errors, alerts, and performance data of your New Relic account. This guide leads from choosing the region to the first query.

Step 1: Check the prerequisites

You need a New Relic account and an MCP client that supports remote servers. According to the documentation, Claude Code, Codex, Claude Desktop, VS Code, and Kiro CLI are among the supported clients; for Claude Code and Codex the page names minimum versions.

Choose the region

The address depends on the region of your account: https://mcp.newrelic.com/mcp/ for the US, https://mcp.eu.newrelic.com/mcp/ for the EU, and https://mcp.jp.newrelic.com/mcp/ for Japan. A wrong region means your data is not found.

Step 2: Set up in Claude Code

Run claude mcp add --transport http newrelic https://mcp.newrelic.com/mcp/, using the EU address in the EU. On first use the browser opens for the OAuth sign-in. The documentation explicitly recommends OAuth because some tools are only available with it.

claude mcp add --transport http newrelic https://mcp.newrelic.com/mcp/

Other clients

For Claude Desktop the documentation describes the route via the mcp-remote proxy, which requires Node.js. In VS Code you register the HTTP server in the mcp.json file.

Restrict tools

With the HTTP header include-tags you limit the tools by categories such as discovery, alerting, and data-access. This is useful when the assistant should only read data and not generate reports.

Step 3: First queries

Start with simple questions: "Which accounts can I see?", "Which alerts are open?", or "Show the most frequent error groups of the last hour." Check generated NRQL queries before you trust them.

Permissions and sensitive data

The assistant acts on behalf of your user. Use an account with minimal rights. Logs can contain personal data and enter the AI client context.

OAuth via the browser (recommended); some tools are available only via OAuth. API keys are documented only for individual clients.

Access to telemetry, alerts, dashboards, and entities of the account within the role of the signed-in user; the preview tools are read-only.

The server runs hosted by New Relic in the chosen region. Retrieved monitoring data enters the AI client context and therefore reaches the model provider.

Exposure of logs containing personal data or credentials to the model provider, overly broad user permissions, and choosing the wrong region.