Know if it's working, why it failed, and what it cost — all in plain English. Set up yourself in 3 minutes.
Trace every span, attribute every failure, version every prompt. One SDK or any OTel exporter — self-hosted or air-gapped.
Sign up, open our step-by-step wizard, pick your platform. We show you exactly where to click. Done in 3 minutes.
Most tools just collect logs. Trace-lit tells you what happened, why it failed, what it cost, and how to make it better.
From a single span to a full production eval pipeline — every stage in one platform.
Every agent step, AI call, and action — live. Know exactly what your agent did and in what order.
Spans, DAG topology, parent-child nesting — SDK or OTel. Every call captured, every token counted.
When something breaks, Trace-lit tells you which step caused it and what else it affected — in plain English.
Root-cause span detected, cascade identified. Blast radius visualised on the DAG before you roll back.
Every update to your AI's instructions is saved automatically. Compare any two versions side by side.
Every prompt commit hashed, diffed, and linked to the spans it produced. Roll back in one click.
Bookmark real runs as good or bad with one click. Build a library of examples your team can learn from.
Tag spans directly from the trace view. Curate production examples into structured datasets for fine-tuning or eval.
Before a change goes live, Trace-lit automatically checks if it's better or worse than what you had before.
CI quality gate runs your dataset against the new prompt. Fails the build if score drops. No regression ships.
n8n, Flowise, Vercel AI, or custom Python — Trace-lit connects in the background. Nothing changes in how your AI runs.
SDK wrappers for every major framework. Native OTel ingestion for everything else. One endpoint, all traces.
When an agent fails, Trace-lit identifies exactly which step caused it and what else broke as a result — no digging through logs.
Root-cause span flagged, cascade spans marked. Webhook to Slack, PagerDuty, or any endpoint.
See the real dollar cost of every agent run, broken down by step. No surprise bills — just a clear dashboard.
Every change your team makes to the AI's instructions is saved and dated automatically. Compare any two versions with a click.
Every prompt version hashed and stored. Diff any two versions. See which spans each version produced.
Bookmark runs as good or bad directly from the dashboard. Your team builds a shared library of what works — no extra tools.
Tag spans as pass/fail from the trace view. Export as JSONL for fine-tuning or feed directly into the eval engine.
Before any AI change goes live, Trace-lit scores it against your saved examples. If it's worse, the release is blocked automatically.
pytest-style eval runner. Scores against your curated dataset in CI. Returns exit code 1 if quality regresses.
Instead of cryptic error codes, get a one-sentence explanation of what went wrong — written for humans, not engineers.
See every message your AI sent and received. Enterprise automatically redacts names, cards, and sensitive data before it's stored.
Enterprise cloud keeps all data in your region. Self-hosted option available for regulated industries.
One docker compose up. ReplicatedMergeTree ClickHouse, Kafka, TimescaleDB. Nothing phones home.
We manage everything. You just use the dashboard. No servers, no setup complexity.
MANAGED BY TRACE-LIT · NO INSTALLATION · PROXY SETUP IN 2 MINUTES
Your servers, your data, your rules. Open source forever — no usage caps, no data sharing, no lock-in.
YOUR SERVERS · YOUR DATA · FULL CONTROL · NEVER LEAVES YOUR INFRASTRUCTURE
docker compose upNo coding required. Sign up free, follow the wizard, and your dashboard is live in 3 minutes.
Open source forever. Self-host on your own infrastructure or use the cloud. Enterprise support when you need it.