For AI teams shipping agents to production

LLM Observability That Finds What's Breaking

Detect failure patterns automatically and discover more via semantic search. Tracked from new to resolved. OTEL-compatible. Free plan, no credit card.

Free plan: 20K credits/month No credit card required Setup in under 5 minutes

Used by engineers at these companies

GitHubNvidiaAWSShopifyDiscordRevolutFiverrWixKayakGoDaddyCognizant

Setup in minutes

3 lines. Any stack. Traces flowing.

OTEL-compatible SDK for TypeScript and Python. Works alongside Datadog and Sentry. Pick your provider below.

Ready to go. Free plan, no credit card.

LLM Providers

Agent Frameworks

How it works

From raw traces to resolved issues

Three screens. That is the entire workflow.

01

Capture everything

Every agent interaction traced automatically. Spans, sessions, tool calls, multi-turn conversations. Full context, not just request/response pairs.

Latitude traces dashboard showing agent spans, latency, and cost metrics
02

Failures cluster into issues

Some failure patterns detected automatically; discover more by searching traces semantically and flagging failures. Named, prioritized by frequency, tracked from New to Escalating to Resolved. A prioritized issue queue.

Latitude issues dashboard showing failure patterns with lifecycle status
03

Review, resolve, detect regressions

Human verdicts on traces with context. Fix it, deploy. If the pattern comes back, Latitude resurfaces the regression automatically.

Latitude issue detail view showing lifecycle tracking from New to Resolved to Regressed
Every failure improves the next detection. Continuous loop.

Honest comparison

How Latitude compares

All open source. All trace LLM calls. Here is where they differ.

Feature
Langfuse
Braintrust
Helicone
Latitude
Observability
Tracing
Full tracing
Full tracing + Brainstore
Proxy-based request logging
Full tracing
Multi-turn agent sessions
Sessions + agent graphs
Session-level tracing
Session grouping via headers
Full conversation context + tool calls
OTEL compatibility
OTEL-native SDKs
Native OTLP + exporters
OpenLLMetry for async
TS + Python + any OTEL
Issue Management
Issue discovery
Semi-automated semantic grouping
ML-based topic clustering
Manual filtering only
Auto-detected + user-driven via semantic search
Issue lifecycle
No issue entity
No formal lifecycle
Not available
New / Escalating / Resolved / Regressed
Regression detection
CI/CD + score alerting
CI/CD gates + drift alerts
Experiments + threshold alerts
Auto-surfaces after deploy
Eval generation from failures
Manual failure → dataset
Loop agent assists
Not available
Auto-generated on issue creation
Evaluations
Evaluations
LLM-as-a-judge
LLM-as-a-judge + custom
LLM-as-a-judge + human feedback
LLM-as-a-judge + custom

All open source. All trace LLM calls. The difference: Latitude turns raw observability into a closed-loop improvement system.

Already using Datadog or Sentry? Latitude runs alongside it. OTEL-compatible, same protocol you already use. Latitude handles the AI-specific layer: failure patterns, issue lifecycle, agent session tracing. Your APM handles infrastructure.

4,000+

GitHub stars

1,200+

Community members

MIT

Open source

Self-host

Your infrastructure

Teams using Latitude in production

Pew Research CenterSuperlistPlannedLegalitasRetracedVirtuous

Pricing

Start free. Scale when you need to.

Unlimited seats on every plan. No per-user pricing.

Starter

Free
  • 20K credits/month
  • Unlimited seats
  • 30-day retention
  • Full observability + evals
Most popular

Pro

$99 /month
  • 100K credits/month
  • Unlimited seats
  • 90-day retention
  • $20 per 10K extra credits

Enterprise

Custom
  • Custom volume
  • Custom retention
  • SAML SSO, on-prem
  • SOC2, ISO 27001

Stop hunting through logs

See what's failing. Fix it systematically. Free plan, no credit card required.

Free plan: 20K credits/month No credit card required Setup in under 5 minutes