No more pattern-hunting in traces
Individual trace review does not reveal failure patterns at scale. Engine clusters failures automatically so you see what is broken and how often, without digging through thousands of traces.
Now Available
Stop manually triaging agent failures. Let Engine fix it.
LangSmith Engine watches your production traces, clusters failures into named issues, diagnoses root causes, and proposes fixes and eval coverage automatically.
Available now in public beta. No credit card required.
Get started with LangSmith Engine in three steps
Connect a tracing project
Link a LangSmith tracing project. Engine immediately starts monitoring your production traces for failure signals across errors, evaluator results, latency anomalies, and user feedback.
Optionally connect your repo
Give Engine access to your codebase and it can read relevant code, diagnose root causes, and draft targeted PRs with fixes ready for your review.
Review and merge improvements
Engine surfaces clustered issues in a prioritized list, each with a proposed fix, a custom online evaluator, and failing traces added to your offline eval dataset. You review and ship.
LangSmith powers top engineering teams, from AI startups to global enterprises
.svg)
.svg)
.svg)
.svg)
.svg)
Built on the trusted agent engineering platform
LangSmith Engine uses the traces, evals, and development context teams already capture in LangSmith to help them improve agents without standing up new infrastructure.
50M+
LLM Calls Traced Daily
1B+
Events Ingested Per Day
100K+
Monthly active orgs in LangSmith
Engine runs the full improvement cycle for you
From clustered failure detection to drafted fixes and eval coverage, Engine handles manual triage work that takes most teams hours.
Engine monitors traces for explicit errors, evaluator failures, latency spikes, negative user feedback, and unusual behavior, finding both known and undetected problems. When it detects a repeated pattern, it clusters those traces into a named issue with severity, timeline, and evidence links.
Connect with our team to see how
Built on LangSmith Infrastructure
No new infrastructure to manage
LangSmith Engine is powered by your existing LangSmith setup. It reads from your current tracing projects and evaluator results, and writes back into your existing offline eval workflow.
Connects to your repo
Link your repository and Engine can read relevant code to diagnose root causes and open targeted PRs directly against your codebase.
SOC 2 Type II
LangSmith is SOC 2 Type II certified with comprehensive security controls for enterprise teams.
Trust centerWhy teams are switching to Engine for agent improvement
Fixes and evals together
Engine proposes a code or prompt fix and an evaluator for the same issue at the same time. When you merge the fix, the evaluator is already in place to catch regressions.
Plugs into your existing setup
Engine works on top of your current LangSmith tracing projects and evaluator results. Connect a project, optionally connect your repo, and it starts surfacing issues immediately.
What early Engine users are saying
"Engine saves our team hours of digging by not only identifying emerging failure modes, but also proactively suggesting evals and code changes to resolve them quickly."
"It has dramatically reduced our time-to-detection and time-to-triage, meaning less time spent hunting through traces and a much shorter path from broken behavior to a deployed fix."
See LangSmith Engine in action
Talk to our team about how Engine can automate failure detection, root cause diagnosis, and eval coverage for your agents. 20-minute demo.