LIBRA
Watching production nowLives in your SlackQueryable over MCP

Production broke at 3:07 AM.You get paged.Libra already detected the incident.

By 3:14 it had read the traces, named the deploy that caused it, and opened the fix PR with a regression test. Libra watches your logs, traces, and deploys around the clock. You review and merge in the morning.

Sign in
01 / 05Detect03:07:41

Northwind Freight and Brightvale Health can't send replies, ~40 queued and stalled. Libra leads with the customer impact in Slack itself, before anyone is paged.

anomaly detected03:07:41
# prod-alertsSlack
Libraapp3:07 AM
prod p1 · customer impact

2 workspaces can't send follow-up replies right now.

POST /api/followups/send is failing: 14 errors in 8 min, 0 the hour before.

affected workspaces
NWNorthwind Freight24 replies stalled
BVBrightvale Health16 replies stalled
Sending workflow only. Everything else healthy.

Posted proactively. Nobody paged it, nobody typed @libra.

Swipe or tap through the night.

Agents that watch prod 24/7.

Production failures aren't straightforward.

An alert tells you that there's a spike in error rates. A customer says something stopped working. The last deployment completed successfully, but the product has quietly started behaving differently.

Somebody still has to work out what happened, who it affected, what to rollback, and how to verify the fix.

Libra's AI agents start that work before your team has to reconstruct everything by hand.

Works with the stack you already have.

Our agents use existing tooling and monitoring systems your team already has to perform investigations.

  • Logs and tracesEvery request your service handled, and where it spent its time
    • Datadog
    • Sentry
    • Grafana
  • Deploys and pull requestsWhat changed, when it shipped, which lines it touched
    • GitHub
    • GitLab
  • Cloud infrastructureRestarts, memory, exit codes, the machines underneath
    • Amazon Web Services
    • Microsoft Azure
    • Google Cloud
  • Product analyticsWhat people clicked, and where the page gave up on them
    • PostHog
    • Amplitude
  • Team conversationsWhat your team already tried, ruled out, or happens to know
    • Slack
    • Microsoft Teams
  • Incident historyPast incidents, and every correction someone gave it
    • PagerDuty
    • incident.io
  • Custom data sourcesInternal tools, business systems, and signals unique to your product

What Libra investigates

  • Where the failure starts in the request path: service, API, queue, database or external dependency
  • Which deployment, configuration or infrastructure change aligns with the first bad signal
  • The blast radius across customer accounts, workflows and integrations
  • Whether the system is still degrading, stable or recovering against its normal baseline
  1. An alert in Slack/TeamsDetails on the blast radius, affected services, APIs, tenants and downstream dependencies. Reply in-thread to ask Libra questions
  2. An evidence-backed diagnosisContaining the correlated traces, logs, metrics and recent changes supporting the hypothesis
  3. Suggested operational actionsRollback, configuration change, owner escalation or further diagnostic step
  4. Recovery verificationMonitor production to verify the original errors and dependent service health have returned to normal
Libra reads logs and traces, deploys and pull requests, cloud infrastructure, product analytics, team conversations, past incidents, and custom data sources. Libra investigates where a failure starts, which recent change aligns with it, its blast radius, and whether the system is degrading, stable, or recovering against its normal baseline. It then produces an alert in Slack or Teams, an evidence-backed diagnosis, suggested operational actions, and recovery verification.

Who we are

Libra is built by a group of Amazon and Apple engineers, backed by Y Combinator. It's the product we wanted to exist back when we were the ones on call. Reach out and we'll get back to you within the hour (PT).

Watch an incident response, start to finish

How we compare.

Detects incidents without pre-built alerts

Datadog detects on its own. Sentry auto-catches errors and slowdowns. Resolve wakes on deploys and schedules; incident.io needs a declared incident. Libra watches raw telemetry and catches failures that have no monitor yet.

Libradoes this
Datadog~Sentryincident.io~Resolve.ai

Learns what normal looks like for your business

Others learn normal metric curves. Libra learns your business: which customers are active, what each endpoint normally does. A dormant test account spiking stays out of your pages.

Libradoes this
~DatadogSentryincident.ioResolve.ai

Verifies the fix against production behavior

Datadog's agent iterates a fix through CI. Sentry drafts tests without running them. Passing tests prove the tests pass. Libra runs the fix beside a baseline in a sandbox and replays the real failing requests from your traces when they cover the change.

Libradoes this
~DatadogSentryincident.ioResolve.ai

Verifies recovery in production before closing

Datadog validates rollouts and auto-resolves monitors. Sentry reopens if the error returns. Resolve can watch a deploy. Only Libra refuses to close until the exact failure stays quiet in production.

Libradoes this
~Datadog~Sentryincident.io~Resolve.ai

Blocks root-cause claims that lack evidence

Others attach evidence and confidence to findings, and Datadog can mark an investigation inconclusive; labeling a guess is still posting it. Libra fails closed: no evidence in the message, no message.

Libradoes this
~DatadogSentry~incident.io~Resolve.ai
does this~ a slice of it not today