> ## Documentation Index
> Fetch the complete documentation index at: https://www.adaline.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# With Adaline Integrations

> Connect Adaline to popular AI frameworks, providers, and OpenTelemetry-based applications

Use Adaline Integrations when your application already runs through an AI framework, provider SDK, gateway, or OpenTelemetry pipeline. Integrations let you keep your existing application structure while sending logs, traces, spans, model calls, tool calls, retrieval steps, and agent activity into Adaline.

This path is best when you want to observe production traffic without hand-building every trace and span yourself.

## When to use integrations

| Use case                                                                                                                                                           | Start here                                                                                             |
| ------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------ |
| You use LangChain, LangGraph, LlamaIndex, DSPy, LiteLLM, Pydantic AI, AutoGen, CrewAI, Mastra, Vercel AI SDK, or OpenTelemetry.                                    | [Framework integrations](/integrations/introduction)                                                   |
| You want to connect provider traffic from OpenAI, Anthropic, Google, Vertex AI, Azure, Bedrock, Groq, OpenRouter, Together AI, xAI, or another supported provider. | [AI provider integrations](/integrations/introduction)                                                 |
| You want the fastest provider-compatible logging path by changing the base URL.                                                                                    | [With Adaline Proxy](/instrument/with-adaline-proxy)                                                   |
| You need complete control over custom traces, spans, sessions, variables, and metadata.                                                                            | [With Adaline SDKs](/instrument/with-adaline-sdks) or [With Adaline API](/instrument/with-adaline-api) |

## What integrations should capture

A good integration sends enough information for Adaline to understand what happened and improve the agent later:

* Trace and span names that describe the workflow.
* Model requests and responses.
* Tool calls, arguments, responses, and status.
* Retrieval, embedding, guardrail, command, or custom workflow spans when available.
* Latency, token usage, cost, model, and provider details.
* Session, environment, release, route, prompt, agent, and customer-safe metadata.
* User feedback or application outcome when you can safely send it.

## Browse integrations

<CardGroup cols={2}>
  <Card title="All integrations" icon="puzzle" href="/integrations/introduction">
    Browse Adaline integrations for AI providers, frameworks, gateways, agents, and OpenTelemetry.
  </Card>

  <Card title="OpenTelemetry" icon="radio-tower" href="/integrations/frameworks/opentelemetry">
    Send OTLP traces from OpenTelemetry-compatible applications into Adaline.
  </Card>

  <Card title="LangChain" icon="https://mintcdn.com/adaline/Um_T8BffW4hfcoYD/images/icons/langchain.svg?fit=max&auto=format&n=Um_T8BffW4hfcoYD&q=85&s=0133af6154fe31fefda23420a6478ace" href="/integrations/frameworks/langchain" width="24" height="24" data-path="images/icons/langchain.svg">
    Observe LangChain chains, tools, retrieval, and model calls in Adaline.
  </Card>

  <Card title="LangGraph" icon="https://mintcdn.com/adaline/Um_T8BffW4hfcoYD/images/icons/langgraph.svg?fit=max&auto=format&n=Um_T8BffW4hfcoYD&q=85&s=1d6dd52f67db7747f003fd49c84c25b0" href="/integrations/frameworks/langgraph" width="24" height="24" data-path="images/icons/langgraph.svg">
    Capture stateful graph execution, agent steps, and tool use.
  </Card>

  <Card title="Vercel AI SDK" icon="https://mintcdn.com/adaline/asQy_HFY3PYbBCEt/images/icons/vercel-ai.svg?fit=max&auto=format&n=asQy_HFY3PYbBCEt&q=85&s=bc57c08fe97ca9f2cac62016bfca1ca7" href="/integrations/frameworks/vercel-ai" width="76" height="65" data-path="images/icons/vercel-ai.svg">
    Observe generateText, streamText, tool calls, and multi-step AI SDK flows.
  </Card>

  <Card title="LiteLLM" icon="https://mintcdn.com/adaline/asQy_HFY3PYbBCEt/images/icons/litellm.svg?fit=max&auto=format&n=asQy_HFY3PYbBCEt&q=85&s=c85c55824791ea3f146e5fe3245a53a3" href="/integrations/frameworks/litellm" width="64" height="64" data-path="images/icons/litellm.svg">
    Attach Adaline while keeping your existing LiteLLM provider routing.
  </Card>
</CardGroup>

## Verify the integration

After wiring an integration, run one known request in staging and confirm:

1. The request appears in [Logs](/monitor/analyze-log-traces).
2. The trace has readable spans for model calls, tools, retrieval, or agent steps.
3. The trace includes environment, route, release, prompt, or agent metadata.
4. Tokens, latency, cost, model, and provider details appear where supported.
5. Sessions or Behaviors become useful once enough production traffic arrives.

If the integration does not expose the level of detail you need, use [With Adaline SDKs](/instrument/with-adaline-sdks), [With Adaline API](/instrument/with-adaline-api), or [Advanced usage](/instrument/advanced-usage) for more control.