New Relic MCP for NRQL, APM, logs, traces, entities, and incident context with account-aware observability retrieval.
Looking for more MCP servers? Browse the full MCP tools directory or explore more tools in Monitoring.
New Relic MCP turns telemetry into an entity graph an agent can traverse: an application entity connects to transactions, errors, distributed traces, logs, deployments, alert conditions, and incidents. That matters because an incident is rarely contained in one chart; the useful question is why did checkout latency rise after release X, and answering it requires joining deployment metadata to span timing, error rate, and infrastructure signals. The hidden architectural problem is account and region identity. New Relic accounts, regions, API endpoints, and query permissions are valid independently, so a successful NRQL request can still target the wrong account or an incomplete data plane. Preview status and changing tool surfaces add another operational variable. Use a read-only user key, explicit region and account ID, bounded NRQL time windows, and query-result limits. Require the agent to identify the entity GUID and data timestamp before interpreting an empty result. New Relic MCP is strongest for application-performance correlation and weakest when treated as a generic log search box.
Choose New Relic MCP when you want an MCP server focused on Monitoring and need tighter integration with your existing tools.
Good fit
Use the following configuration as a starting point for Claude Desktop or any compatible MCP client, then replace placeholder credentials with your own values.
{
"mcpServers": {
"new-relic": {
"url": "https://mcp.newrelic.com/mcp"
}
}
}These setup steps cover the typical installation flow for this MCP server.
Enable the New Relic AI MCP Server preview in Administration > Previews & Trials if your account requires it.
Connect the hosted endpoint and complete OAuth or API authorization for the correct New Relic region.
Pro-Tip: The fatal flaw 90% of developers commit is starting with a global NRQL query and trusting an empty result. Account IDs, entity GUIDs, EU/US region endpoints, and retention windows can all produce valid emptiness. Require the agent to report account ID, region, entity GUID, query window, and event type before diagnosing health.
Use a read-only user key for analysis and keep mutation, alert, and workflow permissions outside the default agent identity.
Start from an entity or incident, then narrow to transactions, spans, logs, and deployment events; cap results before expanding the time range.
Verify with a known application entity, a short NRQL query, and an incident or deployment correlation.
Common questions for New Relic MCP.
The silent failure is account, region, entity, or retention mismatch. NRQL executed correctly but queried a different account, wrong regional endpoint, wrong event type, or a time range outside retention. The engineering fix is to print account ID, region, entity GUID, event type, and time window before treating empty data as healthy.
Both support observability workflows, but New Relic is especially strong around entity, APM, NRQL, and distributed application context; Datadog offers an integrated metrics, logs, traces, and monitor ecosystem. Choose the platform containing your operational source of truth.
Availability and capabilities depend on the New Relic preview and account configuration. Verify current preview terms, supported clients, and tool permissions before making it an incident-critical dependency.
Only when permissions and tools allow it. Keep the default identity read-only and require explicit approval for alert mutations, deployments, and remediation.
| Feature | New Relic MCP | Competitor |
|---|---|---|
| Application context | ✅ Entity, APM, NRQL, traces, logs, and deployments | ❌ Standalone log search without application topology |
| Query language | ✅ NRQL with account-aware telemetry | ✅ Vendor-specific query language and metric model |
| Deployment maturity | ❌ Hosted MCP capability may depend on preview status | ✅ Mature connector availability varies by platform |
| Incident correlation | ✅ Strong entity-centric correlation | ❌ Depends on configured integrations and tags |