Understand your system.

How Lantern works

Lantern receives traces, metrics and logs over OTLP/HTTP. The browser and read-only MCP tools query the same authenticated application API. Your instrumentation decides what is recorded.

The Go ingestion gateway validates each batch and stores original per-record Protobuf envelopes alongside typed fields in GreptimeDB. Accounts use PostgreSQL. These databases run on the Lantern side; connecting an application does not require installing them.

A span becomes a wide event: one operation with its name, timing, status and exported attributes. Filter matching spans, group them by another field, inspect a trace or compare slow operations with the rest.

Choose a starting point

What you need

• Access: a Lantern account for the browser and an ingest key for exporters, which a workspace administrator creates under Account settings. AI assistants sign in with your account.
• Telemetry: OpenTelemetry instrumentation or an existing Collector. An endpoint setting alone does not create spans, host metrics or logs.
• Connectivity: outbound HTTPS to the ingestion origin; MCP also needs HTTPS access to the app origin.

You can export directly from a compatible SDK or forward data from a Collector. There is no Lantern-specific application SDK to install. See the requirements and dependency table.

Read the evidence carefully

Queries describe observed telemetry. Sampling, missing instrumentation, export failures, retention and query limits can leave gaps. A span count is not automatically a request count, and a visible root does not prove a complete trace.

Lantern exposes incomplete and truncated results instead of filling gaps with demo data. Keep warnings and the exact time window with any conclusion. Correlation can guide the next investigation; it does not establish a cause.

The public pilot currently retains telemetry for seven days. Production capacity, complete Collector compatibility and replacement parity with other observability products remain workload-specific validation tasks.

Continue with traces, metrics and logs to understand what each signal can tell you.