Skip to content
Drill

Something's wrong in production.Drill finds out why.

Ask your cluster what's wrong. Drill investigates logs, metrics, traces and your code, shows you every step, and comes back with the root cause.

  • Runs in your own cluster
  • Never changes your cluster
  • Bring your own model

A replay of a Drill investigation: it plans four steps, runs 36 tool calls across Kubernetes, logs, metrics and traces, and answers: checkout-api isn't broken. The error count is driven by 400 validation failures on startDate from one client sending a malformed date since 10:11 UTC — errors went from ~1,000/h to ~20,000/h. No 5xx surge, no crash-looping, no resource pressure.

Your on-call'sfavorite on-call.

See every step

The plan and every tool call, live, with the raw output behind each conclusion.

Watches your cluster for you

Checks run on their own and tell you what to fix first.

Survives a closed laptop

Investigations run in the background and resume where they stopped.

Remembers every fix

Solved incidents come back as citations the next time.

Nothing risky runs without you

Sensitive commands wait for a person to approve them.

From “something's wrong”to a fix you trust.

01

Ask your cluster

Describe the symptom the way you'd tell a colleague. Get the cause, the fix and how to check it.

How investigations work
02

Ask your repositories

Connect GitHub or Bitbucket and a production error is traced to the function behind it.

How code awareness works
03

Never solve it twice

A solved investigation becomes a reviewed write-up that every later answer can cite.

How knowledge builds up
04

Catch it before it pages you

Scheduled checks and Alertmanager alerts, matched to the workload they're about.

How monitoring works

Connect once,investigate everything.

  • Kubernetes
    Workloads, events, rollouts, control plane and etcd
  • Prometheus
    The metrics behind a symptom
  • Alertmanager
    Alerts investigated as they fire
  • Grafana
    The dashboards already watching a service
  • Grafana Loki
    The log lines that explain an error
  • Grafana Tempo
    A request followed across services
  • GitHub · Bitbucket
    Production errors traced to the code
  • PostgreSQL
    Backups, WAL archiving and wraparound
  • MySQL
    Durability settings and replication
  • MongoDB
    Oplog window and replica-set health
  • ClickHouse
    Part counts, tiny inserts, stuck mutations
  • RabbitMQ
    Memory alarms and queues nobody consumes

Your cluster, your data.Read-only by design.

  • Runs in your cluster

    Installed with Helm. No telemetry, no phone-home.

  • Never changes your cluster

    A read-only role; fixes come back as diffs.

  • Bring your own model

    Local, Azure AI Foundry, Bedrock in your region or Vertex AI.

How Drill stays safe

Stop guessing.Drill it.