> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agentlinkops.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Monitor placements and inspect changes

> Follow source and destination checks, interpret event feeds, pause recurring work, and review usage without confusing unknown attempts with loss.

A hosted active watch keeps a supplied placement available for scheduled checks. Start with [check a placement](/guides/check-placement), or explicitly enroll a [verified candidate](/guides/import-and-verify). Local CLI checks run only when invoked.

## Review the latest evidence

Call [list\_link\_watches](/reference/commands/list_link_watches) for a compact list. Open [get\_link\_watch](/reference/commands/get_link_watch) for the selected record and [get\_link\_history](/reference/commands/get_link_history) for its observations. A recent attempt can be unknown while an older complete observation remains the last known state. Keep both dates in your report.

Use [list\_link\_events](/reference/commands/list_link_events) to consume changes. Apply a complete page to your local records before storing `next_cursor`. Use event identifiers to avoid duplicate effects during replay.

Destination health has its own watches, history and [event feed](/reference/commands/list_target_events). Maintain its cursor separately. An unavailable destination does not establish loss of the backlink that cites it.

## Control recurring work

Use [update\_link\_watch](/reference/commands/update_link_watch) to pause a placement or change its expectations and cadence. Pausing preserves history. Cadence is a requested interval; workspace limits and scheduling affect when work runs. Read [get\_usage](/reference/commands/get_usage) to distinguish reserved work from consumed and released check units. Source and destination checks share allowance.

Request an extra check only when the task needs new evidence. Reuse the request's idempotency value after a transport failure. A delayed job needs inspection, not repeated new submissions.

## Recover a missed interval

For `CURSOR_EXPIRED`, follow the returned snapshot recovery instructions. Save the recovered snapshot before resuming from its cursor; do not treat a missing retained interval as a quiet period. The [sync workflow](/guides/sync-and-export) handles this for the local ledger.

Set up [webhooks](/guides/webhooks) to notify a receiver of changes when delivery is configured. Read [evidence semantics](/evidence) before reporting confirmed loss, and use [troubleshooting](/guides/troubleshooting) for pending checks or delivery failures.