Skip to main content
Use New Relic through Corsair: one client, typed API calls. Use a user key, not a license key. Create one at https://one.newrelic.com/api-keys under Create a key, choosing User as the key type. A license key only ingests data and cannot read any of this. The plugin talks to two surfaces, both under the same key:
  • REST v2 for applications, metrics and alerts, at api.newrelic.com/v2.
  • NerdGraph, the GraphQL API, for NRQL, accounts, entities, issues and deployment markers, at api.newrelic.com/graphql.
An EU account uses api.eu.newrelic.com for both, and a key is only valid in its own region. Pass region: 'eu' on any call from an EU account; the default is us. What to know:
  • health_status is a colour, such as green, orange, red, gray or unknown. The REST API says orange where the UI shows yellow, and unknown means the app has no alert policy attached.
  • reporting goes false after ten quiet minutes, and last_reported_at is then an empty string rather than a timestamp, so a missing time is not an error.
  • A metric window defaults to the last 30 minutes when from is omitted. summarize: true collapses the window into one value per metric, which is what most dashboards want.
  • REST v2 pages through a Link header, which this plugin cannot read, so page is an input. A long list needs an explicit page walk.
  • alerts.conditions covers APM, browser and mobile conditions only. NRQL conditions, which most accounts use, are alerts.nrqlConditions and its create, update and delete.
  • Updates replace the whole object. alerts.updatePolicy needs both name and incident_preference, and alerts.updateNrqlCondition needs the whole condition: New Relic answers 422, 500 or 400 to a partial body. Read the current values first.
  • Find GUIDs with entities.search. It takes an entity filter such as type = 'APPLICATION', domain = 'APM' or name LIKE 'checkout', and pages with nextCursor.
  • issues.list is the current incident model (NerdGraph aiIssues). title, description and a few other fields are lists even when they hold one value.
  • A deployment marker cannot be deleted. changeTracking.createDeployment writes it on the entity’s charts for good, so it is destructive: the permission policy decides whether it needs approval (cautious asks, strict refuses, open runs it).
  • A NRQL query is a NerdGraph call. accountId is required, and the answer carries the time window New Relic actually ran over, which can differ from the SINCE clause.
  • A bad NRQL query answers 200 with an errors array. The plugin turns that into a thrown error rather than returning rows that are silently empty.
  • Deleting a policy deletes its conditions, and New Relic has no restore call.
  • incident_preference is one of PER_POLICY, PER_CONDITION or PER_CONDITION_AND_TARGET and decides whether alerts roll up per policy, per condition, or per condition and signal. It defaults to PER_POLICY.
What you get:
  • 18 typed API operations

Setup

1

Install

2

Add the plugin

corsair.ts
With multiTenancy: true, scope calls with corsair.withTenant(id). See Quick start for KEK + Hub keys, and Multi-tenancy for account isolation.
3

Choose authentication

4

Connect a tenant

Mint a connect link and send the tenant to it. Hub hosts the page and delivers the result to your app — see Connect / OAuth.

Example API calls

List applications with their health
Create an alert policy
See the full list on the API page.

What’s next

API reference

Every newrelic.api.* operation with input and output types.

Connect / OAuth

createLink, Hub delivery, and tenant connect flows.

Use with agents

Expose this plugin’s operations as MCP tools.