Observability for backend teams
Tracefield joins traces, logs and metrics into one timeline — so the 3 a.m. page tells you which service broke, not just that something did.
npm i -g tracefield
v2.4.1 · Free for 3 services · First trace in 90 seconds
Latency, not vibes
p50 tells you the typical request. p99 tells you the one your customer actually remembers. Tracefield plots all three, always.
One timeline
Stop correlating across five tools at 3 a.m. Every span links to its logs; every log links back to its trace.
Every request gets a trace ID at the edge. Follow it through 14 services without guessing which hop was slow.
Log lines carry their trace ID. Click one and you are inside the exact request — not a grep across a day of logs.
Golden signals per service — latency, traffic, errors, saturation — each chart one click from the trace that caused it.
Burn-rate alerts on error budgets, not on every spike. Grouped, deduped, routed to the on-call who owns the service.
One agent per host, auto-instrumentation for the frameworks you already run. No code changes, no redeploy to start seeing traces.
OTLP in, OTLP out. Your data stays portable — export to any backend, any time, with no lock-in migration project.
Trace walkthrough
Trace 9f31…c2 · POST /checkout · 412ms total. The waterfall shows where the time actually went — no guessing, no adding log lines and redeploying.
Two lines of code
The SDK wraps your framework's request handler. Spans, logs and metrics flow automatically from there.
1import { tracefield } from '@tracefield/node'; 2 3// One line. Every route gets a trace, every error gets context. 4tracefield.init({ service: 'checkout-api', sampleRate: '1.0' }); 5 6app.post('/checkout', async (req, res) => { 7 const span = tracefield.startSpan('charge-card'); 8 await charge(req.body); // spans nest automatically 9 span.end(); 10 res.json({ ok: true }); 11});
Pricing
Every plan includes traces, logs and metrics — no feature is held hostage behind the top tier.
For side projects and learning the ropes.
For teams shipping to production every day.
For platforms with real traffic and real SLAs.
FAQ
One agent per host, plus auto-instrumentation for Node, Python, Go and Java frameworks. No code changes to start; add two lines when you want custom spans.
Measured overhead is under 1% of request latency at 100% sampling, and under 0.2% with tail sampling on. The agent batches and compresses before sending — it never blocks your request path.
It is yours. Tracefield speaks OpenTelemetry natively, so you can export everything — traces, metrics, dashboards — to any OTLP-compatible backend at any time. No export fees, no hostage formats.
Yes. Open-source projects get the Team plan free — apply with your repository URL and we approve within two working days.
A real outage, dissected span by span. No marketing, no spam.