New Relic MCP Server – Observability AI Access
New Relic's MCP server lets AI development tools query observability data, investigate alerts, analyze performance, and retrieve context from New Relic accounts using the Model Context Protocol.
Overview
New Relic MCP is the official Model Context Protocol integration for connecting AI development environments to New Relic observability data. It is designed for teams that want an AI agent to answer operational questions with live telemetry context instead of relying on static knowledge. The server is currently documented by New Relic as a preview feature and is configured against New Relic's hosted MCP endpoint.
What the MCP server enables
The server exposes specialized tools for observability and incident analysis. AI agents can discover accounts and entities, fetch dashboards, run NRQL queries, translate natural-language requests into NRQL, inspect alert policies and recent issues, list synthetic monitors, retrieve recent logs, analyze deployments, review error groups, and summarize user impact. New Relic organizes these tools with internal tags such as discovery, data-access, alerting, incident-response, performance-analytics, and advanced-analysis; clients can use the include-tags HTTP header to limit the tool set presented to an agent.
When to use it
Use this MCP server when an AI assistant needs to investigate production behavior in New Relic. Practical workflows include asking which entities exist in an account, checking recent alerts for a service, looking up top errors from the last hour, running NRQL against NRDB, analyzing whether a deployment affected performance, or gathering context for an incident review. It is most useful for SRE, platform, observability, and engineering teams that already use New Relic as their telemetry and alerting system.
Connection and authentication
New Relic documents a hosted HTTP MCP endpoint at https://mcp.newrelic.com/mcp/ for US accounts and a regional endpoint at https://mcp.eu.newrelic.com/mcp/ for EU accounts. OAuth is documented as the recommended method for supported clients. API-key authentication is also supported through an Api-Key or api-key HTTP header using a New Relic user API key. Claude Desktop uses mcp-remote as a client-side proxy for OAuth, but the provider-supported connection remains the hosted HTTP MCP endpoint.
Key considerations
Access is governed by the New Relic user, OAuth profile, or API key configured during setup. If the credential lacks permission to view alerts, execute NRQL, or read an account or entity, the corresponding tool call fails. New Relic explicitly warns users to apply RBAC and least-privilege access because AI tools may act on behalf of the configured user. New Relic also documents compliance restrictions: the public preview MCP server must not be used with FedRAMP- or HIPAA-regulated accounts or data.
Supported Transports
streamable_http
URL: https://mcp.newrelic.com/mcp/
Frequently Asked Questions
- When should an AI agent use the New Relic MCP server?
- Use it when an AI agent needs live observability context from New Relic, such as investigating production alerts, retrieving dashboards, querying telemetry with NRQL, analyzing deployments, or summarizing errors and recent issues for incident response.
- What does New Relic MCP add to an AI agent's capabilities?
- It gives the agent access to New Relic MCP tools for entity discovery, account lookup, dashboard retrieval, NRQL execution, natural-language to NRQL conversion, alert policy inspection, recent issue review, log access, deployment impact analysis, and performance analytics.
- What can the AI agent access through New Relic MCP?
- The agent can access only the New Relic accounts, entities, telemetry, dashboards, alerts, logs, and analysis functions allowed by the configured OAuth user or API key. New Relic RBAC and key permissions determine which tool calls succeed.
- How is authentication configured for New Relic MCP?
- New Relic documents OAuth as the recommended option for supported clients. API-key authentication is also supported by sending a New Relic user API key in an HTTP header such as Api-Key or api-key. Secrets should be stored in environment variables and tied to least-privilege New Relic roles.
- Which transport should be used for New Relic MCP?
- Use the hosted HTTP MCP endpoint. The US endpoint is https://mcp.newrelic.com/mcp/ and New Relic documents https://mcp.eu.newrelic.com/mcp/ for EU accounts. Client-side tools such as mcp-remote may proxy the connection for clients that need it, but the provider endpoint is HTTP.