Pick a target
These are the only supported pairs. Exact detection reports every text difference. Semantic detection confirms a text difference, then asks an AI judge whether it matters under your
instructions. When you omit change_detection, it’s inferred from the target as in the table. Extract monitors cost more per run than page and sitemap monitors. See credits.
Create a monitor
This monitor watches a changelog page and reports only the changes its instructions describe. Because the target hasinstructions, detection is semantic. The examples read your API key from CONTEXT_DEV_API_KEY; the Quickstart shows how to set one up.
201 with the monitor and initial_run_id (trimmed):
initial_run_id to Retrieve a monitor run to confirm the baseline worked. If the run couldn’t be queued, initial_run_id is null and the baseline runs at the next scheduled time.
Store webhook.secret to verify signatures. With no schedule, the monitor runs once a day; see Schedules.
Access
A restricted API key needsmonitors:write to create, update, delete, or run a monitor, rotate its webhook secret, and replay its webhook deliveries. monitors:read covers monitors, runs, changes, limits, credit usage, and delivery history. webhook.secret is returned only to keys with monitors:write. See API keys.
Before you rely on alerts
- The baseline run completed:
status: "completed"andbaseline_created: true. instructionssay what matters and what to ignore.- Your receiver verifies signatures and deduplicates events on
X-Context-Id. - Failed runs send no webhook, so you check runs for failures or subscribe to
run.completedas a heartbeat.
Related
- Targets: page, sitemap, and extract options, detection, and updates.
- Schedules: intervals, run now, pause, and automatic pauses.
- Runs and changes: run statuses and change records.
- Events and alerts: webhook events, delivery failures, and secret rotation.
- Limits and errors: monitor limits per plan and error codes.
- Website change digests: an end-to-end recipe.
- Create a monitor reference