LangChain & LangGraph

LangChain and LangGraph are OpenTelemetry-native through the community instrumentors. Tracelane accepts OTLP directly, so you can configure the exporter yourself as shown below. Alternatively, the Python SDK LangGraph integration wires the same instrumentor through auto_instrument(); choose one setup.

What you get


Agent structure	Yes. The graph, each node, each chain step, prompt / model / parser spans, nested with real parent-child links.
The LLM leg	Yes — model spans with token counts, inside the tree.
Prompt & response content	Only if you enable it in the instrumentor. Tracelane stores what you send.

This is the option that captures the agent, not just the model call. If you instead point your OpenAI-compatible client at the Tracelane gateway base URL, you get the LLM leg only — one span per request, no tree. See Which integration should I use?.

Setup

pip install openinference-instrumentation-langchain \
            opentelemetry-sdk opentelemetry-exporter-otlp-proto-http

from opentelemetry import trace
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.resources import Resource
from opentelemetry.sdk.trace.export import BatchSpanProcessor
from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
from openinference.instrumentation.langchain import LangChainInstrumentor
import os

provider = TracerProvider(resource=Resource.create({"service.name": "my-agent"}))
provider.add_span_processor(BatchSpanProcessor(OTLPSpanExporter(
    endpoint="https://gateway.tracelane.dev/v1/traces",
    headers={"x-tracelane-api-key": os.environ["TRACELANE_API_KEY"]},
)))
trace.set_tracer_provider(provider)

LangChainInstrumentor().instrument(tracer_provider=provider)
# every LangChain and LangGraph run now exports to Tracelane

Your API key needs the ingest scope. Mint one in Settings → API keys.

LangChainInstrumentor covers LangGraph too — LangGraph is built on LangChain runnables, so graph nodes come through the same callback system. You do not need a second instrumentor.

What a trace looks like

A two-node LangGraph agent, verified end to end against Tracelane production on 2026-08-31 — 11 spans, 1 root, 10 parented, 0 orphans:

LangGraph                                    (root)
├── plan
│   └── RunnableSequence
│       ├── ChatPromptTemplate
│       ├── ChatOpenAI
│       └── StrOutputParser
└── act
    └── RunnableSequence
        ├── ChatPromptTemplate
        ├── ChatOpenAI
        └── StrOutputParser

Read back out of Tracelane’s own storage, not asserted from the exporter.

Span volume — read this before you pick a plan

One agent run is many spans. The tree above is 11 spans for one invoke(); a single prompt | model | parser chain is 4. That is the point of the integration — the structure is the product — but it means span volume is not run volume. Plan limits are counted in spans on this path.

An OTel span and a gateway call are not the same unit, and they are not counted the same way. A gateway request and an OTel span both count as ingest bytes against the SAME meter — there is no separate quota for one path or the other. On this OTel path your app sends spans, and one agent run is many spans — the tree above is 11 for a single invoke(). Size your ingest estimate against spans per run, not runs, and see pricing.

Which integration should I use?

	Gateway base URL	OTel exporter (this page)
Setup	change one base URL	~10 lines of exporter config
Captures the LLM call	yes	yes
Captures chain / nodes / tool calls	no	yes
Trace shape	one span per request	a nested tree
Routing, BYOK, guardrails, caching	yes	no — calls go direct to your provider

They compose: point the gateway at your provider and run the instrumentor, and you get gateway features plus agent structure.

The two paths do not currently join into one trace. A gateway-proxied call does not adopt your agent’s trace context, so it appears as its own single-span trace rather than nesting under the LangGraph run. If you need one tree, use the OTel exporter path alone.

Get started

SDKs

Integrations

Concepts

Features

Tools

Audit & compliance

Operations

Migrations

Comparisons

LangChain & LangGraph

What you get

Setup

What a trace looks like

Span volume — read this before you pick a plan

Which integration should I use?

​What you get

​Setup

​What a trace looks like

​Span volume — read this before you pick a plan

​Which integration should I use?

What you get

Setup

What a trace looks like

Span volume — read this before you pick a plan

Which integration should I use?