// integrations · 47 and counting

The stack you already shipped, fully instrumented.

Native SDKs for Node, Python, Go, and Rust. Adapters for every major model provider, vector DB, and agent framework. Pipe traces to your existing APM. No re-platforming, no rebuilds.

$

Model providers

Every major frontier and open-weight provider. Auto-instrumentation captures token counts, model versions, and request IDs.
A
Anthropic
Claude Opus, Sonnet, Haiku. Tool-use, citations, prompt caching all captured.
O
OpenAI
GPT-4.1, o-series, function calling, structured outputs, streaming chunks.
G
Gemini
Gemini Pro, Flash, Ultra. Multimodal inputs traced as separate spans.
B
AWS Bedrock
Drop-in for Claude, Llama, Mistral via Bedrock. IAM-aware.
A
Azure OpenAI
Same auto-instrumentation as OpenAI, with Azure deployment IDs propagated.
C
Cohere
Command R+, Embed v3, rerank API. Captures search and rerank scores.
M
Mistral
Mistral Large, Codestral, embedding endpoints. EU-hosted spans supported.
↯
Replicate
Open-weight model hosting — captures cold-start, queue, and inference.
x
xAI Grok
Grok-3 and Grok-4. Tool use traced; live-search results captured as spans.Beta

Agent frameworks

If you build with one of these, our SDK auto-instruments the framework's call graph end-to-end. No wrapping required.
L
LangChain
Chains, agents, tools, retrievers. v0.2+ supported.
↯
LlamaIndex
Query engines, retrievers, response synthesizers — all instrumented.
V
Vercel AI SDK
streamText, generateText, useChat. Streams traced chunk-by-chunk.
L
LangGraph
Stateful graph execution with per-node spans and state snapshots.
P
Pydantic AI
Type-safe agents in Python; tool schemas captured automatically.
M
Mastra
Workflow-first TypeScript framework; durable steps traced as spans.
C
CrewAI
Multi-agent orchestration with per-agent and per-task tracing.
A
AutoGen
Microsoft's multi-agent conversation framework, fully instrumented.

Vector databases

Captures query latency, top-K, score distribution, and the exact filters applied — so retrieval debugging stops being a black box.
P
Pinecone
Index name, namespace, top-K, hybrid scores all captured per query.
W
Weaviate
GraphQL queries traced; vector and BM25 scores recorded.
Q
Qdrant
Self-hosted and cloud. Filters and rerank stages traced as sub-spans.
C
Chroma
Local and server modes; collection and embedding model captured.
M
MongoDB Atlas
Atlas Vector Search; aggregation pipeline traced.
P
pgvector
Postgres extension; query plans and HNSW params recorded.

Observability & alerting

Pipe Your Business spans into your existing APM. Send eval-regression alerts to where your on-call already looks.
⊳
OpenTelemetry
Native OTel — every span follows semantic conventions. OTLP export.
D
Datadog
Forward traces and metrics; correlate with HTTP and infra spans.
G
Grafana Tempo
OTLP-compatible Tempo backend; LogQL/PromQL on LLM metrics.
H
Honeycomb
Wide-event tracing; rich span attributes survive the trip.
N
New Relic
OTel exporter; agent traces side-by-side with APM transactions.
S
Slack
Eval-regression and budget alerts to channels with severity routing.
P
PagerDuty
Critical eval drops and ingest failures page on-call directly.
// hands-on

Three integrations, three lines.

A real snippet from a real customer codebase, for the three most-installed adapters. No "just configure this YAML" hand-waving.

A

Anthropic

One-line auto-instrumentation. Captures every Claude call across your codebase.

import { instrument } from
  "@yourbusiness/anthropic";

instrument(new Anthropic({ apiKey }));

// from now on, every .messages.create()
// emits a span with full prompt + tokens.
Read full Anthropic guide →
L

LangGraph

Wrap your graph runner. Per-node, per-edge spans with state snapshots.

import { withTracing } from
  "@yourbusiness/langgraph";

const graph = withTracing(myGraph, {
  snapshotState: true,
  name: "qualify_workflow"
});
Read full LangGraph guide →
P

Pinecone

Drop-in factory. Captures top-K, scores, namespace, and filters per query.

import { traced } from
  "@yourbusiness/pinecone";

const index = traced(
  pinecone.Index("sales-leads")
);
Read full Pinecone guide →
// bring your own

If your stack isn't here yet

Your Business is OpenTelemetry-native. Anything that emits OTLP can write to us. Use the manual SDK to instrument private services or proprietary tools — most teams ship a custom integration in under an hour.

// custom span for an internal tool
import { span } from "@yourbusiness/sdk";

await span("internal.crm_sync", async ({ tag }) => {
  tag({ tool: "salesforce", op: "upsert_lead", lead_id });
  return await sf.upsert(lead);
});

// span lands in Your Business alongside your LLM traces,
// fully searchable + filterable + alertable.
// missing one?

Request an integration

If we're missing an adapter you need, tell us. We ship roughly two new integrations a month and we prioritize by request volume — every vote moves it up the queue.

Already shipping with one of these? Connect it now.

Free up to 50,000 spans a month. Most integrations install in under three minutes — no credit card.

Start free → No credit card $ npm i @yourbusiness/sdk