Skip to main content

Documentation Index

Fetch the complete documentation index at: https://sedataai.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

class TelemetryManager implements ObservabilityInstance
TelemetryManager is the implementation behind the ObservabilityInstance you get back from instrumentServer. Most users never construct one directly — instrumentServer does it for you. Reach for it when you want more control: multiple managers per process, advanced testing, or non-MCP instrumentation.

​
Constructor

new TelemetryManager(config: TelemetryConfig)
Validates the config, builds the OpenTelemetry NodeSDK, attaches resource detectors, and starts the SDK.

​
Public properties

​
tracer
Tracer
The OpenTelemetry tracer scoped to your serverName / serverVersion.
​
meter
Meter
The OpenTelemetry meter scoped to your serverName / serverVersion.

​
Methods

​
startActiveSpan(name, attributes, fn)

Open an active span. Adds mcp.session.id to the attributes automatically. 
manager.startActiveSpan('cache.lookup', { 'cache.key': 'user:42' }, (span) => {
  // ...
  span.end()
})

​
createSpan(name, attributes)

Create a span without making it active. Useful when you need a manually managed span. 
const span = manager.createSpan('background.task', {})
// later
span.end()

​
getHistogram(name, options)

Returns a recorder function (value, attributes?) => void. Reuse the same recorder across calls to avoid re-creating the instrument: 
const recordDuration = manager.getHistogram('cache.lookup.duration', {
  description: 'Cache lookup duration',
  unit: 'ms',
})

recordDuration(12, { 'cache.hit': true })

​
getIncrementCounter(name, options)

Returns an increment function (value, attributes?) => void: 
const incRetry = manager.getIncrementCounter('upstream.retry.count', {
  description: 'Upstream retries',
})

incRetry(1, { upstream: 'github' })

​
getArgumentAttributes(params, prefix?)

Flattens an object of tool arguments into a flat attribute map for use in span attributes. Honors enableArgumentCollection — returns {} when off. 
manager.getArgumentAttributes({ city: 'sf', units: 'metric' })
// → { 'mcp.request.argument.city': 'sf', 'mcp.request.argument.units': 'metric' }

​
processTelemetryAttributes(data)

Runs data through every configured dataProcessor in order. 
const cleaned = manager.processTelemetryAttributes({ 'user.email': 'jane@acme.com' })

​
shutdown(): Promise<void>

Records the session-duration histogram and shuts down the underlying NodeSDK. Always call this on graceful exit.

​
Example: standalone use

import { TelemetryManager } from '@sedata-ai/mcp'

const manager = new TelemetryManager({
  serverName: 'background-worker',
  serverVersion: '1.0.0',
  exporterEndpoint: 'https://otel.sedata-ai.tech/v1',
  exporterAuth: { type: 'bearer', token: process.env.SEDATA_TOKEN! },
})

manager.startActiveSpan('worker.run', { 'worker.id': '42' }, async (span) => {
  await doWork()
  span.end()
})

await manager.shutdown()

​
See also

instrumentServer

Higher-level entry point most users want.

Custom instrumentation

Patterns for using the tracer and meter directly.