Newrelic CLI

What this skill does

Translates natural-language observability questions into newrelic CLI commands, resolving account names to numeric IDs automatically. It knows every New Relic account, which environment and service slot each belongs to, and which region (US/EU) to use.

Account resolution

A bundled accounts.json (sibling to this file) maps every New Relic account. Read it at the start of any task to resolve accounts.

Resolution rules

1. If the user names an environment (e.g. "production", "test", "uk-prod"), match against the environment field.
2. If they name a slot (e.g. "services", "envoy", "cardinal"), match against the slot field within the target environment.
3. If they give a numeric account ID, use it directly.
4. If they name an alias (e.g. "Acme_Corp_1", "My_Org"), match against the alias field.
5. Defaults when no account is specified:
   • Service-specific queries (transactions, throughput, errors) -> $DEFAULT_SERVICES_ACCOUNT
   • General queries (infrastructure, custom events) -> $DEFAULT_MAIN_ACCOUNT

Always confirm which account you resolved to before running the query, e.g.: "Querying Services Production ($ACCOUNT_ID)..."

CLI command reference

NRQL queries

The most common operation. Always include --accountId and default to --format JSON for structured output.

newrelic nrql query --accountId <ID> --query '<NRQL>' --format JSON

Example patterns:

# Transaction throughput
newrelic nrql query --accountId $ACCOUNT_ID \
  --query "SELECT rate(count(*), 1 minute) FROM Transaction WHERE appName = 'my-service' TIMESERIES SINCE 1 hour ago" \
  --format JSON

# Error rate
newrelic nrql query --accountId $ACCOUNT_ID \
  --query "SELECT percentage(count(*), WHERE error IS true) FROM Transaction WHERE appName = 'my-service' SINCE 1 hour ago" \
  --format JSON

# P99 latency
newrelic nrql query --accountId $ACCOUNT_ID \
  --query "SELECT percentile(duration, 99) FROM Transaction WHERE appName = 'my-service' TIMESERIES SINCE 1 hour ago" \
  --format JSON

Entity search

Find applications, hosts, containers, and other entities.

newrelic entity search --name <NAME>
newrelic entity search --name <NAME> --domain APM --type APPLICATION
newrelic entity search --tag "environment:production"

Useful filters: --domain (APM, INFRA, BROWSER, SYNTH, etc.), --type (APPLICATION, HOST, CONTAINER, etc.), --tag, --alert-severity, --reporting.

APM operations

# Search for an application
newrelic apm application search --accountId <ID> --name <APP_NAME>

# Record a deployment
newrelic apm deployment create --applicationId <APP_ID> --revision <SHA>

Profile management

Profiles store API keys and default account IDs for repeated use.

newrelic profile list
newrelic profile add --name <PROFILE> --apiKey <KEY> --region <US|EU> --accountId <ID>
newrelic profile default --name <PROFILE>

NerdGraph (GraphQL)

For advanced queries not covered by built-in commands.

newrelic nerdgraph query '{ actor { user { email } } }'

Custom events

newrelic events post --accountId <ID> --event '{"eventType":"Deployment","revision":"abc123"}'

Environment variables

Workflow

1. Parse intent: What does the user want to know or do?
2. Resolve account: Read accounts.json, match environment/slot/alias.
3. Build command: Construct the newrelic CLI invocation.
4. Show the user: Display the resolved account and command before running.
5. Execute: Run via Bash and return results.
6. Interpret: Summarize the data in plain language -- highlight anomalies, trends, or notable values.

Common NRQL patterns

These patterns cover the most frequent observability questions:

-- Service health overview
SELECT count(*), average(duration), percentage(count(*), WHERE error IS true)
FROM Transaction WHERE appName = '{service}'
SINCE 1 hour ago

-- Top slow transactions
SELECT average(duration), count(*)
FROM Transaction WHERE appName = '{service}'
FACET name SINCE 1 hour ago LIMIT 10

-- Error breakdown
SELECT count(*) FROM TransactionError
WHERE appName = '{service}' FACET error.class SINCE 1 hour ago

-- Infrastructure (CPU/memory) -- use main account
SELECT average(cpuPercent), average(memoryUsedPercent)
FROM SystemSample WHERE hostname LIKE '%{service}%'
SINCE 1 hour ago TIMESERIES

-- Kafka consumer lag -- use services account
SELECT average(consumer.totalLag)
FROM KafkaOffsetSample WHERE consumerGroup LIKE '%{service}%'
FACET topic SINCE 1 hour ago

Multi-account queries

When a question spans multiple accounts (e.g. "compare prod vs test error rates"), run separate queries against each account and present the results side by side. Label each result clearly with the account name and environment.

Error handling

• If newrelic is not installed, suggest: brew install newrelic-cli
• If authentication fails, check newrelic profile list and suggest newrelic profile add
• If an account ID is ambiguous (e.g. "services" without an environment), ask the user to clarify which environment