What's covered in each module?

Services: operation-level latency/errors, dependencies, DB/queue breakdowns. Applications (RUM): real-user sessions, errors, and performance for web and mobile apps. Databases: per-datastore throughput, errors, and latency derived from trace spans. Kubernetes: deployments/pods/nodes, restarts, resource hot-spots. Jobs: successes/failures, retries, durations, queue/consumer health. Hosts: CPU/mem/disk/net/process and app↔host correlation.

Know what you run. Fix what matters.

Q: Does it work with AI agents?

Yes. With the Last9 MCP , agents like Claude/Cursor can query Last9 (e.g., "slowest payments ops last 15m") and get RCA reports and fix recommendations.

Auto-discover everything you run and trace problems to root cause—fast. Debug production issues with AI directly in your IDE.

Book demo

Use Discover to show causes, not just symptoms

CONTEXT TO ACT

Discover goes deeper than "service is slow" with ops, DB calls, exceptions, all correlated with telemetry.

CONVENTION FIRST

Beyond service catalogs. Always up-to-date inventory of services, clusters, jobs, and hosts.

ONE-CLICK PIVOTS

Jump from alerts to correlated logs, metrics, and traces without changing context or tools.

AI ASSISTED

Use Last9 with Claude, Cursor, or Slack to get hypothesis and RCA reports along with fix recommendations.

For the needs of modern-day micro-services architectures.

Discover Services

Know every service, its health, dependencies, and user impact.
Operation analytics: latency, errors, throughput, saturation.
Exceptions & slow queries with drill-downs.
Outgoing calls + DB/Cache breakdowns. Instant pivots to related logs & traces.
Auto-mapped upstream/downstream dependencies.

Discover Kubernetes

See deployments, pods, resources the way you actually run them.
Live view of deployments, pods, nodes. CPU / memory / network / storage utilization.
Pod-level errors and restarts; workload hot-spots.
See deployment → pod health with CPU/mem/network and restart patterns for fast isolation.

Discover Jobs

Background & async, finally visible.
Execution success/failures, retries, and durations.
Queue/consumer breakdowns and throughput trends.
Exception hotspots and payload-level context.
Related traces/logs for stuck or poison messages.

Discover Hosts

When it's not the code, it's the box.
CPU, memory, disk, network, and process metrics.
Noisy-neighbor detection and saturation alerts.
Host → service/process correlation for quick isolation.

Discover Databases

Every datastore your services query, with nothing to install on the database server.
Throughput, error rate, and P95 latency per database and per operation.
Slowest and most frequent queries, each linked to its traces.
Top services by load: find the one consumer before you scale the instance.

Agents Monitoring

What your LLM calls cost, how long they take, and which ones fail.
Cost per conversation, total tokens, and P50 to P99 latency.
Every slow or failing call links to its full request trace.
Top models and providers actually serving traffic.

Last9 helped us forget all the different observability tools and consolidate every dashboard into one single place.

Rahul Mahale

Principal DevOps Engineer

Frequently asked questions

Can't find the answer to your question?

Talk to our support team for help.

Contact support

A real-time inventory and dependency map for Services, Applications (RUM), Databases, Kubernetes, Jobs, and Hosts that stitches metrics, logs, and traces into one debug surface.

APM says "service is slow." Discover pinpoints which operation, which dependency/pod/host, and deep-links to the exact logs/traces.

OpenTelemetry/Prometheus. Point existing collectors/exporters at Last9 — no proprietary agents.

Yes. It auto-maps what it sees and gets richer as you add coverage.

Minutes — services, pods, jobs, and hosts populate with out-of-the-box dashboards and dependency graphs.

One-click pivots from an alert → impacted operation → related traces/logs → affected pods/hosts for explainable, step-by-step triage.

Services: operation-level latency/errors, dependencies, DB/queue breakdowns.
Applications (RUM): real-user sessions, errors, and performance for web and mobile apps.
Databases: per-datastore throughput, errors, and latency derived from trace spans.
Kubernetes: deployments/pods/nodes, restarts, resource hot-spots.
Jobs: successes/failures, retries, durations, queue/consumer health.
Hosts: CPU/mem/disk/net/process and app↔host correlation.

Yes — mirror telemetry and compare side-by-side. No lock-in.

Yes. With the Last9 MCP, agents like Claude/Cursor can query Last9 (e.g., "slowest payments ops last 15m") and get RCA reports and fix recommendations.

No extra runtime beyond your OTel instrumentation. Backend is built for high-cardinality and bursty traffic. We've handled peaks of 1.2B events/min and 60M+ concurrent users.