v2.4 · Streaming traces are live

Observability that speaks agent.

Your Business is a tracing and eval platform built for LLM-native applications. Drop in three lines of code and get distributed traces, token-cost attribution, and live evals across every model, tool call, and retry — without the cardinality bill.

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

Trusted by Cursor, Linear, Notion AI, Replicate, and 4,200+ teams shipping with LLMs.

~/projects/sales-copilot · live trace

$ yourbusiness trace --follow agent-7c4e
# streaming live trace from production…

● root qualify_lead 204ms
├─ tool crm.fetch_account 62ms · cache HIT
├─ llm gpt-4.1 / structured 132ms · 1,204 tok · $0.0042
├─ eval factuality_v3 PASS · 0.91
└─ tool slack.notify 10ms

● root draft_followup 3.4s
├─ llm claude-opus-4.7 2.1s · 4,820 tok · $0.072
├─ retry tool_call_failed → 1 retry · ok
└─ eval brand_voice PASS · 0.94

# 2 traces · p95 latency 3.6s · spend $0.094 · 0 errors

// what yourbusiness does

Three lines of SDK. Every layer of your agent, observed.

We instrument the entire LLM call graph — prompts, tool calls, retries, and evals — so you can ship agents the way you ship the rest of your product: with traces, dashboards, and on-call.

⊳

Distributed tracing

OTel-compatible spans for every LLM call, tool call, and retry — across services. Search, filter, replay.

import { trace } from "@yourbusiness/sdk";
// wraps any async function
await trace("qualify", async () => { ... });

⊕

Live evals

Factuality, helpfulness, brand-voice. Run a panel of evals on every production trace, not just CI.

yourbusiness.eval("factuality_v3", {
  on: "every_response",
  alert: { score: "< 0.85" }
});

Cost attribution

Token spend, broken down by feature, customer, model, and retry. Stop being surprised by the bill.

yourbusiness.tag(span, {
  customer_id: "acme",
  feature: "summarize"
});

★

Prompt registry

Version, A/B, and roll back prompts without a deploy. Every change is a diffable artifact.

const prompt = await
  yourbusiness.prompt("qualify@latest");
// canary 10% to qualify@v18

∿

Replay & debug

Replay any production trace locally with the same inputs, model, and tools. Reproduce bugs in 30 seconds.

$ yourbusiness replay trc_a1b2c3
→ rebuilding tool sandbox…
→ replaying with claude-opus-4.7

↯

Datasets from prod

Curate eval datasets from real traces. Click to add, version-control, run regression suites in CI.

yourbusiness.dataset("qualify-edge")
  .add(trace)
  .freeze("v3");

// pricing

Priced per span, not per seat.

Free for hobby. Pro scales linearly with traces, no surprise cardinality bills. Enterprise is for teams that need SSO, audit log, and a signed BAA.

// Hobby

Free

$0/ forever

For solo builders and side projects. Forever free, no credit card.

50,000 spans / month
3 days trace retention
1 project · 1 user
Community Discord

// Pro

Pro

$49/ month per project

For teams shipping LLM features in production. Linear pricing, no per-seat tax.

5M spans included · $0.10 per 1k after
30 days retention · replay + datasets
Unlimited seats · roles + audit log
Live evals (5 panels included)
Slack + PagerDuty alerting

// Enterprise

Talk to us

Custom

For regulated teams and 100+ engineer orgs. Self-host or VPC-deploy.

SSO / SAML · SCIM provisioning
Self-hosted or in-VPC deploy
SOC 2 · HIPAA · signed BAA
Volume pricing · dedicated CSM
99.99% uptime SLA