Docs

Shows direct mainnet RPC observations with provider labels and UTC check times. Slot polling refreshes every 10 seconds. A head difference is not a time-based finality measurement.

Enter a slot, a Base58 block hash, or a transaction signature. Blocks show parent slot, hashes, block time and transaction signatures; choose a signature to inspect status and fee. Processed, confirmed and finalized states are kept separate. A processed slot may not yet have an available block. Blockhash search covers the latest 32 finalized slots; use a slot for older blocks. Missing or pruned data stays unavailable.

A scheduled server collector runs every minute, starts at finalized root + 1, and correlates first confirmed-block observation with first finalized observation of the identical slot and blockhash. Durations use a local monotonic clock within one run. Only matched blocks become samples; duplicates are idempotent within a run. These are polling lower bounds, not full consensus latency. Gaps and unmatched runs are shown, not replaced with zero. RPC time is plotted separately. The approximately 150 ms Alpenglow target is never presented as measured finality. Stored observations remain after you close the page. Scheduled collection uses the published version; no observations are fabricated before it runs.

Notarization certificates, finalization certificates, round events, stake thresholds and BLS verification remain unavailable. Standard public RPC does not expose the required messages. Operational inspection needs a real native Alpenglow feed or instrumented validator, original signatures, signer keys and validator-set context. The certificate API returns 503 until such a source exists.

Filter by identity or vote account and choose active or delinquent accounts. Select an identity to inspect epoch vote credits and commission. Status, activated stake and last vote come directly from getVoteAccounts. Stake shares and top-ten concentration are calculated from those observations. Epoch credits are not a participation percentage. Scheduled snapshots retain vote-account history every 15 minutes; the stored snapshot table shows active/delinquent counts and stake concentration over time. An expected-vote denominator is not available.

Cross-point latency remains unavailable until independent instrumented observation points are connected with synchronized clocks, uncertainty records and stable block identifiers. RPC response differences are not propagation. Rotor-specific telemetry requires a real exposed Rotor implementation.

Available endpoints cover health, summary, slots, finality and validators under /api/public/. No wallet is required. Input validation and a local per-instance request limit protect public reads. Native certificates are explicitly non-operational. Dedicated WebSocket/Geyser ingestion workers, durable streaming reconnects, global distributed rate limiting and full archival indexing require additional infrastructure; this MVP uses bounded RPC polling.

Separates current request connectivity from scheduled collection history. A stale or failed collector is a gap, not proof of a network outage. No continuous uptime percentage is inferred from occasional checks.