Microsoft OpenTelemetry Distro

Prerequisites: Python 3.10 or later.

pip install microsoft-opentelemetry

Prerequisites: Node.js 18 or later. ESM apps require Node.js 20.6.0 or later.

npm install @microsoft/opentelemetry

node --import @microsoft/opentelemetry/loader ./dist/index.js

Install the Microsoft.OpenTelemetry NuGet package. This single package includes the Agent 365 observability runtime, hosting integration, auto-instrumentation for supported AI frameworks, and standard OpenTelemetry libraries.

<PackageReference Include="Microsoft.OpenTelemetry" />

Or use the .NET CLI:

dotnet add package Microsoft.OpenTelemetry

Call use_microsoft_opentelemetry() to enable observability.

from microsoft.opentelemetry import use_microsoft_opentelemetry
from microsoft.opentelemetry.a365.hosting.token_cache_helpers import AgenticTokenCache

token_cache = AgenticTokenCache()

use_microsoft_opentelemetry(
    enable_a365=True,
    a365_token_resolver=lambda agent_id, tenant_id: (
        (t := asyncio.run(token_cache.get_observability_token(agent_id, tenant_id)))
        and t.token or None
    ),
)

For custom token resolution (instead of the default token resolver), see Manual token resolver.

You can customize the exporter behavior by passing optional a365_* kwargs to use_microsoft_opentelemetry().

Use ObservabilityManager with AgenticTokenCacheInstance to handle token caching automatically:

import { useMicrosoftOpenTelemetry, AgenticTokenCacheInstance } from '@microsoft/opentelemetry';
import { resourceFromAttributes } from '@opentelemetry/resources';

useMicrosoftOpenTelemetry({
  resource: resourceFromAttributes({
    'service.name': 'Default Token Cache Sample',
    'service.version': '1.0.0',
  }),
  azureMonitor: {
    enabled: Boolean(process.env.APPLICATIONINSIGHTS_CONNECTION_STRING),
  },
  instrumentationOptions: {
    http: { enabled: false },
  },
  a365: {
    enabled: process.env.ENABLE_A365_OBSERVABILITY_EXPORTER !== 'false',
    tokenResolver: (agentId, tenantId) =>
      AgenticTokenCacheInstance.getObservabilityToken(agentId, tenantId) ?? '',
  },
});

For custom token resolution (instead of the default token resolver), see Manual token resolver.

You can customize the exporter behavior by setting optional properties on the a365 object.

For full configuration options and environment variables, see the JavaScript repository README.

In Program.cs, call UseMicrosoftOpenTelemetry() to enable observability. The ExportTarget flags enum controls where telemetry is sent - combine targets with | (for example, ExportTarget.Console | ExportTarget.Agent365).

Auto (DI) token cache (recommended for Agent Framework apps):

When you don't set TokenResolver, the distro automatically registers IExporterTokenCache<AgenticTokenStruct> via DI. Your agent calls RegisterObservability() at runtime to supply credentials, and the cache handles token acquisition and refresh.

// Program.cs — no TokenResolver needed
builder.UseMicrosoftOpenTelemetry(o =>
{
    o.Exporters = ExportTarget.Agent365;
    // TokenResolver is auto-registered via the agentic token cache in DI, with required imports in the next code snippet.
});

using Microsoft.Agents.A365.Observability.Hosting.Caching;
using Microsoft.Agents.A365.Observability.Runtime.Common;

// In your agent class
public class MyAgent : AgentApplication
{
    private readonly IExporterTokenCache<AgenticTokenStruct> _agentTokenCache;

    public MyAgent(AgentApplicationOptions options,
        IExporterTokenCache<AgenticTokenStruct> agentTokenCache) : base(options)
    {
        _agentTokenCache = agentTokenCache;
    }

    protected async Task MessageActivityAsync(
        ITurnContext turnContext, ITurnState turnState, CancellationToken cancellationToken)
    {
        _agentTokenCache.RegisterObservability(
            turnContext.Activity.Recipient.AgenticAppId,
            turnContext.Activity.Recipient.TenantId,
            new AgenticTokenStruct(
                userAuthorization: UserAuthorization,
                turnContext: turnContext,
                authHandlerName: "AGENTIC"),
            EnvironmentUtils.GetObservabilityAuthenticationScope());
    }
}

For custom token resolution (instead of the default token resolver), see Manual token resolver.

You can customize the exporter behavior by setting optional properties on o.Agent365.Exporter.

from microsoft.opentelemetry.a365.core import BaggageBuilder

with (
    BaggageBuilder()
    .tenant_id("tenant-123")
    .agent_id("agent-456")
    .conversation_id("conv-789")
    .build()
):
    # Any spans started in this context will receive these as attributes
    pass

To auto-populate the BaggageBuilder from the TurnContext, use the populate helper in the microsoft-opentelemetry package. This helper automatically extracts caller, agent, tenant, channel, and conversation details from the activity.

from microsoft.opentelemetry.a365.core import BaggageBuilder
from microsoft.opentelemetry.a365.hosting.scope_helpers.populate_baggage import populate

builder = BaggageBuilder()
populate(builder, turn_context)

with builder.build():
    # Baggage is auto-populated from the TurnContext activity
    pass

import { BaggageBuilder, InvokeAgentScope, ExecuteToolScope } from '@microsoft/opentelemetry';

// Create and apply baggage context
const baggageScope = new BaggageBuilder()
  // Core identifiers
  .tenantId('tenant-123')
  .agentId('agent-456')
  .conversationId('conv-789')
  .build();

// Execute operations within the baggage context
baggageScope.run(() => {
  // All spans created within this context will inherit the baggage values

  // Invoke another agent
  const agentScope = InvokeAgentScope.start(request, scopeDetails, agentDetails);
  // ... agent logic

  // Execute tools
  const toolScope = ExecuteToolScope.start(request, toolDetails, agentDetails);
  // ... tool logic
});

To auto-populate the BaggageBuilder from the TurnContext, use the fromTurnContext helper in the @microsoft/opentelemetry package. This helper automatically extracts caller, agent, tenant, channel, and conversation details from the activity.

import { BaggageBuilder, BaggageBuilderUtils } from '@microsoft/opentelemetry';

const baggageScope = BaggageBuilderUtils.fromTurnContext(new BaggageBuilder(), context)
  .invokeAgentServer(context.activity.serviceUrl, 3978)
  .build();

baggageScope.run(() => {
  // Baggage is auto-populated from the TurnContext activity
});

using Microsoft.Agents.A365.Observability.Runtime.Common;

using var baggageScope = new BaggageBuilder()
    .TenantId("tenant-123")
    .AgentId("agent-456")
    .ConversationId("conv-789")
    .Build();
// Any spans started in this context will receive them as attributes.

To auto-populate the BaggageBuilder from the ITurnContext, use the FromTurnContext extension method in the Microsoft.Agents.A365.Observability.Hosting package. This method automatically extracts caller, agent, tenant, channel, and conversation details from the activity.

using Microsoft.Agents.A365.Observability.Runtime.Common;
using Microsoft.Agents.A365.Observability.Hosting.Extensions;

using var baggageScope = new BaggageBuilder()
    .FromTurnContext(turnContext)
    .Build();

In Python, register baggage middleware through ObservabilityHostingManager.configure() rather than directly on the adapter.

from microsoft.opentelemetry.a365.hosting import ObservabilityHostingManager, ObservabilityHostingOptions

options = ObservabilityHostingOptions(enable_baggage=True)
ObservabilityHostingManager.configure(adapter.middleware_set, options)

The middleware skips baggage setup for async replies (ContinueConversation events) to avoid overwriting baggage that the originating request already set.

Register BaggageMiddleware on the adapter. It automatically extracts caller, agent, tenant, channel, and conversation details from every incoming TurnContext and wraps the request in a baggage scope.

import { BaggageMiddleware } from '@microsoft/opentelemetry';

// Option 1: Register middleware directly on the adapter
adapter.use(new BaggageMiddleware());

Alternatively, use ObservabilityHostingManager to configure baggage middleware along with other hosting features.

import { ObservabilityHostingManager } from '@microsoft/opentelemetry';

const manager = new ObservabilityHostingManager();
manager.configure(adapter, { enableBaggage: true });

The middleware skips baggage setup for async replies (ContinueConversation events) to avoid overwriting baggage that the originating request already set.

Register BaggageTurnMiddleware on the adapter. It automatically extracts caller, agent, tenant, channel, and conversation details from every incoming ITurnContext and wraps the request in a baggage scope.

using Microsoft.Agents.A365.Observability.Hosting.Middleware;

adapter.Use(new BaggageTurnMiddleware());

The middleware skips baggage setup for async replies (ContinueConversation events) to avoid overwriting baggage that the originating request already set.

If you need HTTP-level baggage (for example, to set tenant and agent IDs before the Bot Framework pipeline runs), register ObservabilityBaggageMiddleware in the ASP.NET Core pipeline by using the UseObservabilityRequestContext extension method. You must provide a resolver function that extracts the tenant ID and agent ID from the HTTP context.

using Microsoft.Agents.A365.Observability.Hosting.Middleware;

app.UseObservabilityRequestContext((httpContext) =>
{
    // Extract tenant and agent IDs from your request context
    var tenantId = GetTenantIdFromContext(httpContext);
    var agentId = GetAgentIdFromContext(httpContext);
    return (tenantId, agentId);
});

from microsoft.opentelemetry import use_microsoft_opentelemetry

def token_resolver(agent_id, tenant_id):
    return "your-token"

use_microsoft_opentelemetry(
    enable_a365=True,
    a365_token_resolver=token_resolver,
    instrumentation_options={
        "semantic_kernel": {"enabled": True},
    },
)

Semantic Kernel auto-instrumentation isn't supported for Node.js.

Semantic Kernel instrumentation is enabled by default. To disable it, set the instrumentation option explicitly.

builder.UseMicrosoftOpenTelemetry(o =>
{
    o.Instrumentation.EnableSemanticKernelInstrumentation = false;
});

from microsoft.opentelemetry import use_microsoft_opentelemetry

def token_resolver(agent_id, tenant_id):
    return "your-token"

use_microsoft_opentelemetry(
    enable_a365=True,
    a365_token_resolver=token_resolver,
    instrumentation_options={
        "openai_agents": {"enabled": True},
    },
)

import { useMicrosoftOpenTelemetry } from "@microsoft/opentelemetry";

const tokenResolver = (agentId, tenantId) => {
  return "your-token";
};

useMicrosoftOpenTelemetry({
  a365: {
    enabled: true,
    tokenResolver,
  },
  instrumentationOptions: {
    openaiAgents: { enabled: true },
  },
});

OpenAI and Azure OpenAI instrumentation is enabled by default. To disable it, set the instrumentation option explicitly.

builder.UseMicrosoftOpenTelemetry(o =>
{
    o.Instrumentation.EnableOpenAIInstrumentation = false;
});

from microsoft.opentelemetry import use_microsoft_opentelemetry

def token_resolver(agent_id, tenant_id):
    return "your-token"

use_microsoft_opentelemetry(
    enable_a365=True,
    a365_token_resolver=token_resolver,
    instrumentation_options={
        "agent_framework": {"enabled": True},
    },
)

Agent Framework auto-instrumentation isn't supported for Node.js.

Agent Framework instrumentation is enabled by default. To disable it, set the instrumentation option explicitly.

builder.UseMicrosoftOpenTelemetry(o =>
{
    o.Instrumentation.EnableAgentFrameworkInstrumentation = false;
});

from microsoft.opentelemetry import use_microsoft_opentelemetry

def token_resolver(agent_id, tenant_id):
    return "your-token"

use_microsoft_opentelemetry(
    enable_a365=True,
    a365_token_resolver=token_resolver,
    instrumentation_options={
        "langchain": {"enabled": True},
    },
)

import { useMicrosoftOpenTelemetry } from "@microsoft/opentelemetry";

const tokenResolver = (agentId, tenantId) => {
  return "your-token";
};

useMicrosoftOpenTelemetry({
  a365: {
    enabled: true,
    tokenResolver,
  },
  instrumentationOptions: {
    langchain: { enabled: true },
  },
});

LangChain auto-instrumentation isn't listed for .NET.

from microsoft.opentelemetry.a365.core import (
    AgentDetails,
    Channel,
    InvokeAgentScope,
    InvokeAgentScopeDetails,
    Request,
    ServiceEndpoint,
)

agent_details = AgentDetails(
    agent_id="agent-456",
    agent_name="Email Assistant",
    agent_description="An AI agent powered by Azure OpenAI",
    agentic_user_id="auid-123",
    agentic_user_email="agent@contoso.com",
    agent_blueprint_id="blueprint-789",
    tenant_id="tenant-123",
)

request = Request(
    content="Please help me organize my emails",
    session_id="session-42",
    conversation_id="conv-xyz",
    channel=Channel(name="msteams"),
)

scope_details = InvokeAgentScopeDetails(
    endpoint=ServiceEndpoint(hostname="myagent.contoso.com", port=443),
)

with InvokeAgentScope.start(
    request=request,
    scope_details=scope_details,
    agent_details=agent_details,
) as scope:
    scope.record_input_messages(["Please help me organize my emails"])

    # Run the agent invocation.

    invoke_scope.record_output_messages(["I found 15 urgent emails."])

import { InvokeAgentScope } from "@microsoft/opentelemetry";
import type {
  AgentDetails,
  A365Request,
  InvokeAgentScopeDetails,
} from "@microsoft/opentelemetry";

const agentDetails: AgentDetails = {
  agentId: "agent-456",
  agentName: "Email Assistant",
  agentDescription: "An AI agent powered by Azure OpenAI",
  agentAUID: "auid-123",
  agentEmail: "agent@contoso.com",
  agentBlueprintId: "blueprint-789",
  tenantId: "tenant-123",
};

const request: A365Request = {
  content: "Please help me organize my emails",
  sessionId: "session-42",
  conversationId: "conv-xyz",
  channel: { name: "msteams" },
};

const scopeDetails: InvokeAgentScopeDetails = {
  endpoint: { host: "myagent.contoso.com", port: 443 },
};

const invokeScope = InvokeAgentScope.start(request, scopeDetails, agentDetails);

try {
  await invokeScope.withActiveSpanAsync(async () => {
    invokeScope.recordInputMessages(["Please help me organize my emails"]);

    // Run the agent invocation.

    invokeScope.recordOutputMessages(["I found 15 urgent emails."]);
  });
} catch (error) {
  invokeScope.recordError(error as Error);
  throw error;
} finally {
  invokeScope.dispose();
}

using Microsoft.Agents.A365.Observability.Runtime.Tracing.Contracts;
using Microsoft.Agents.A365.Observability.Runtime.Tracing.Scopes;

var agentDetails = new AgentDetails(
    agentId: "agent-456",
    agentName: "Email Assistant",
    agentDescription: "An AI agent powered by Azure OpenAI",
    agenticUserId: "auid-123",
    agenticUserEmail: "agent@contoso.com",
    agentBlueprintId: "blueprint-789",
    tenantId: "tenant-123");

var request = new Request(
    content: "Please help me organize my emails",
    sessionId: "session-42",
    conversationId: "conv-xyz",
    channel: new Channel(name: "msteams"));

var scopeDetails = new InvokeAgentScopeDetails(
    endpoint: new Uri("https://myagent.contoso.com"));

using var invokeScope = InvokeAgentScope.Start(
    request: request,
    scopeDetails: scopeDetails,
    agentDetails: agentDetails);

invokeScope.RecordInputMessages(new[] { "Please help me organize my emails" });

// Run the agent invocation.

invokeScope.RecordOutputMessages(new[] { "I found 15 urgent emails." });

from microsoft.opentelemetry.a365.core import (
    ExecuteToolScope,
    ServiceEndpoint,
    ToolCallDetails,
    ToolType,
)

tool_details = ToolCallDetails(
    tool_name="email-search",
    arguments={"query": "from:manager@contoso.com"},
    tool_call_id="tool-call-456",
    description="Search emails by criteria",
    tool_type=ToolType.FUNCTION.value,
    endpoint=ServiceEndpoint(
        hostname="tools.contoso.com",
        port=8080,
        protocol="https",
    ),
)

with ExecuteToolScope.start(
    request=request,
    details=tool_details,
    agent_details=agent_details,
) as scope:
    result = search_emails(tool_details.arguments)
    scope.record_response(result)

import { ExecuteToolScope } from "@microsoft/opentelemetry";
import type { ToolCallDetails } from "@microsoft/opentelemetry";

const toolDetails: ToolCallDetails = {
  toolName: "email-search",
  arguments: JSON.stringify({ query: "from:manager@contoso.com" }),
  toolCallId: "tool-call-456",
  description: "Search emails by criteria",
  toolType: "function",
  endpoint: {
    host: "tools.contoso.com",
    port: 8080,
    protocol: "https",
  },
};

const scope = ExecuteToolScope.start(request, toolDetails, agentDetails);

try {
  await invokeScope.withActiveSpanAsync(async () => {
    const result = await searchEmails(toolDetails.arguments);
    scope.recordResponse(JSON.stringify(result));
  });
} catch (error) {
  scope.recordError(error as Error);
  throw error;
} finally {
  scope.dispose();
}

var toolDetails = new ToolCallDetails(
    toolName: "email-search",
    arguments: "{\"query\": \"from:manager@contoso.com\"}",
    toolCallId: "tool-call-456",
    description: "Search emails by criteria",
    toolType: "function",
    endpoint: new Uri("https://tools.contoso.com:8080"));

using var scope = ExecuteToolScope.Start(
    request: request,
    details: toolDetails,
    agentDetails: agentDetails);

// Execute the tool.

scope.RecordResponse("{\"count\": 15}");

from microsoft.opentelemetry.a365.core import (
    InferenceCallDetails,
    InferenceOperationType,
    InferenceScope,
)

inference_details = InferenceCallDetails(
    operationName=InferenceOperationType.CHAT,
    model="gpt-4o-mini",
    providerName="azure-openai",
)

with InferenceScope.start(
    request=request,
    details=inference_details,
    agent_details=agent_details,
) as scope:
    scope.record_input_messages(["Summarize the following emails for me."])
    response = call_llm()
    scope.record_output_messages([response.text])
    scope.record_input_tokens(response.usage.input_tokens)
    scope.record_output_tokens(response.usage.output_tokens)
    scope.record_finish_reasons(["stop"])

import { InferenceOperationType, InferenceScope } from "@microsoft/opentelemetry";
import type { InferenceDetails } from "@microsoft/opentelemetry";

const inferenceDetails: InferenceDetails = {
  operationName: InferenceOperationType.CHAT,
  model: "gpt-4o-mini",
  providerName: "azure-openai",
};

const scope = InferenceScope.start(request, inferenceDetails, agentDetails);

try {
  await invokeScope.withActiveSpanAsync(async () => {
    scope.recordInputMessages(["Summarize the following emails for me."]);
    const response = await callLLM();
    scope.recordOutputMessages([response.text]);
    scope.recordInputTokens(response.usage.inputTokens);
    scope.recordOutputTokens(response.usage.outputTokens);
    scope.recordFinishReasons(["stop"]);
  });
} catch (error) {
  scope.recordError(error as Error);
  throw error;
} finally {
  scope.dispose();
}

var inferenceDetails = new InferenceCallDetails(
    operationName: InferenceOperationType.Chat,
    model: "gpt-4o-mini",
    providerName: "Azure OpenAI");

using var scope = InferenceScope.Start(
    request: request,
    details: inferenceDetails,
    agentDetails: agentDetails);

scope.RecordInputMessages(new[] { "Summarize the following emails for me." });

// Call the model.

scope.RecordOutputMessages(new[] { "Here is your email summary." });
scope.RecordInputTokens(145);
scope.RecordOutputTokens(82);
scope.RecordFinishReasons(new[] { "stop" });

from microsoft.opentelemetry.a365.core import OutputScope, Response, SpanDetails

# Capture this before exiting the originating InvokeAgentScope context.
parent_context = invoke_scope.get_span_context()
response = Response(
    messages=["Here is your organized inbox."],
)

with OutputScope.start(
    request=request,
    response=response,
    agent_details=agent_details,
    user_details=None,
    span_details=SpanDetails(parent_context=parent_context),
) as scope:
    pass

import { OutputScope } from "@microsoft/opentelemetry";
import type { A365SpanDetails, OutputResponse } from "@microsoft/opentelemetry";

// Capture this before disposing the originating InvokeAgentScope.
const parentContext = invokeScope.getSpanContext();
const response: OutputResponse = {
  messages: ["Here is your organized inbox."],
};

const scope = OutputScope.start(
  request,
  response,
  agentDetails,
  undefined,
  { parentContext } as A365SpanDetails
);

scope.dispose();

// Capture this before disposing the originating InvokeAgentScope.
var parentContext = invokeScope.GetActivityContext();
var response = new Response(new[] { "Here is your organized inbox." });

using var scope = OutputScope.Start(
    request: request,
    response: response,
    agentDetails: agentDetails,
    spanDetails: new SpanDetails(parentContext: parentContext));

export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317

from microsoft.opentelemetry import use_microsoft_opentelemetry

use_microsoft_opentelemetry()

export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317

import { useMicrosoftOpenTelemetry } from "@microsoft/opentelemetry";

useMicrosoftOpenTelemetry();

builder.UseMicrosoftOpenTelemetry(o =>
{
    o.Exporters = ExportTarget.Otlp;
    o.OtlpEndpoint = new Uri("http://localhost:4317");
});

export ENABLE_A365_OBSERVABILITY_EXPORTER=false

from microsoft.opentelemetry import use_microsoft_opentelemetry

def token_resolver(agent_id, tenant_id):
    return "local-validation-token"

use_microsoft_opentelemetry(
    enable_a365=True,
    a365_token_resolver=token_resolver,
)

# Run instrumented application code.

export ENABLE_A365_OBSERVABILITY_EXPORTER=false

import {
  useMicrosoftOpenTelemetry,
  shutdownMicrosoftOpenTelemetry,
} from "@microsoft/opentelemetry";

const tokenResolver = (agentId, tenantId) => {
  return "local-validation-token";
};

useMicrosoftOpenTelemetry({
  a365: {
    enabled: true,
    tokenResolver,
  },
});

// Run instrumented application code.

await shutdownMicrosoftOpenTelemetry();

builder.UseMicrosoftOpenTelemetry(o =>
{
    o.Exporters = ExportTarget.Console;
});

The resolver must be synchronous. Acquire the token in your async activity handler (or via MSAL) and cache it for the resolver.

from microsoft.opentelemetry import use_microsoft_opentelemetry
from microsoft.opentelemetry.a365.runtime import get_observability_authentication_scope

_cached_token: str | None = None

def my_token_resolver(agent_id: str, tenant_id: str) -> str | None:
    return _cached_token

use_microsoft_opentelemetry(enable_a365=True, a365_token_resolver=my_token_resolver)

@AGENT_APP.activity("message", auth_handlers=["AGENTIC"])
async def on_message(context: TurnContext, _state: TurnState):
    global _cached_token
    _cached_token = await AGENT_APP.auth.exchange_token(
        context,
        scopes=get_observability_authentication_scope(),
        auth_handler_id="AGENTIC",
    )

import { useMicrosoftOpenTelemetry } from "@microsoft/opentelemetry";
import { TurnState, AgentApplication, TurnContext, MemoryStorage } from '@microsoft/agents-hosting';
import { ActivityTypes } from '@microsoft/agents-activity';

useMicrosoftOpenTelemetry({
  a365: {
    enabled: true,
    tokenResolver: (agentId: string, tenantId: string, authScopes?: string[]) =>
      myTokenService.getToken(agentId, tenantId),
  },
});


export class A365Agent extends AgentApplication<TurnState> {
  isApplicationInstalled: boolean = true;
  termsAndConditionsAccepted: boolean = true;
  agentName = 'A365 Agent';
  authHandlerName = 'agentic';

  constructor() {
    const useAgenticAuth = process.env.USE_AGENTIC_AUTH === 'true';

    super({
      storage: new MemoryStorage(),
      ...(useAgenticAuth && {
        authorization: {
          agentic: {
            type: 'agentic',
          }
        }
      })
    });

    this.onActivity(ActivityTypes.Message, async (context: TurnContext, state: TurnState) => {
      await this.handleAgentMessageActivity(context, state);
    });
  }

  async handleAgentMessageActivity(turnContext: TurnContext, _state: TurnState): Promise<void> {
    if (process.env.USE_AGENTIC_AUTH !== 'true') {
      console.log('[A365Agent] Authorization not configured, skipping observability token preload');
      return;
    }

    const agentId = turnContext?.activity?.recipient?.agenticAppId ?? '';
    const tenantId = turnContext?.activity?.recipient?.tenantId ?? '';

    // Acquire auth token via OBO exchange
    const authToken = await this.authorization.exchangeToken(turnContext, 'agentic', {
      scopes: ['api://9b975845-388f-4429-889e-eab1ef63949c/Agent365.Observability.OtelWrite']
    });

    // Cache the token so it can be returned by the token resolver
    myTokenService.set(agentId, tenantId, authToken?.token || '');
  }
}

using Microsoft.OpenTelemetry;

builder.UseMicrosoftOpenTelemetry(o =>
{
    o.Exporters = ExportTarget.Agent365;
    o.Agent365.Exporter.TokenResolver = async (agentId, tenantId) =>
    {
        return await MyTokenService.GetTokenAsync(agentId, tenantId);
    };
});

When you set TokenResolver explicitly, the auto DI token cache isn't registered — your resolver is used instead.

from microsoft.opentelemetry import use_microsoft_opentelemetry
from microsoft.opentelemetry.a365.hosting.token_cache_helpers import AgenticTokenCache, AgenticTokenStruct
from microsoft.opentelemetry.a365.runtime import get_observability_authentication_scope

token_cache = AgenticTokenCache()

_cached_tokens: dict[tuple[str, str], str | None] = {}

# Keep the sync resolver side-effect free; refresh the cache in the async request handler.
def sync_token_resolver(agent_id: str, tenant_id: str) -> str | None:
    return _cached_tokens.get((agent_id, tenant_id))

use_microsoft_opentelemetry(enable_a365=True, a365_token_resolver=sync_token_resolver)

@AGENT_APP.activity("message", auth_handlers=["AGENTIC"])
async def on_message(context: TurnContext, _state: TurnState):
    agent_id = context.activity.recipient.id
    tenant_id = context.activity.recipient.tenant_id
    token_cache.register_observability(
        agent_id=agent_id,
        tenant_id=tenant_id,
        token_generator=AgenticTokenStruct(
            authorization=AGENT_APP.auth,
            turn_context=context,
        ),
        observability_scopes=get_observability_authentication_scope(),
    )
    _cached_tokens[(agent_id, tenant_id)] = await token_cache.get_observability_token(
        agent_id, tenant_id,
    )

Use configureA365Hosting to register the A365 middleware on the adapter. This middleware handles baggage and token plumbing automatically.

import { configureA365Hosting } from "@microsoft/opentelemetry";

configureA365Hosting(adapter, {
  enableBaggage: true,
  enableOutputLogging: false,
});

Set enableOutputLogging: false if response content shouldn't be captured as span attributes.

// Program.cs — no TokenResolver needed
builder.UseMicrosoftOpenTelemetry(o =>
{
    o.Exporters = ExportTarget.Agent365;
    // IExporterTokenCache<AgenticTokenStruct> is auto-registered via DI
});

In your agent class, call RegisterObservability() on each turn:

using Microsoft.Agents.A365.Observability.Hosting.Caching;
using Microsoft.Agents.A365.Observability.Runtime.Common;

public class MyAgent : AgentApplication
{
    private readonly IExporterTokenCache<AgenticTokenStruct> _agentTokenCache;

    public MyAgent(AgentApplicationOptions options,
        IExporterTokenCache<AgenticTokenStruct> agentTokenCache) : base(options)
    {
        _agentTokenCache = agentTokenCache;
    }

    protected async Task MessageActivityAsync(
        ITurnContext turnContext, ITurnState turnState, CancellationToken cancellationToken)
    {
        _agentTokenCache.RegisterObservability(
            turnContext.Activity.Recipient.AgenticAppId,
            turnContext.Activity.Recipient.TenantId,
            new AgenticTokenStruct(
                userAuthorization: UserAuthorization,
                turnContext: turnContext,
                authHandlerName: "AGENTIC"),
            EnvironmentUtils.GetObservabilityAuthenticationScope());
    }
}