bulk-pause

Bulk pause / resume

Preflight — read first. If you cannot see any uptimerobot:* MCP tools in your tool list, invoke the uptimerobot:setup skill before doing anything else. Do not tell the user the MCP is misconfigured — setup's Step 0 detects the common case (server connected, tools loaded after session start) and resolves it without re-keying.

There is no single "pause all" tool. The workflow is: enumerate the target monitors with list-monitors, then call update-monitor-status for each.

Use this when the user says "pause everything", "silence alerts during deploy", "mute the api monitors", or "bring them all back online".

Prefer maintenance-windows for scheduled downtime. Maintenance windows suppress alerts without changing monitor state and don't count against your plan's active-monitor cap on resume. Only bulk-pause when the user explicitly wants the monitors stopped, or when a maintenance window isn't suitable (ad-hoc, very short, or already in the middle of a deploy).

Step 1 — List the target monitors

Filter as tightly as possible to avoid pausing unrelated monitors.

By tag

{ "search": "", "filter": [], "limit": 50 }

list-monitors doesn't support tag-based server-side filtering. Pull the page, then filter client-side on tagNames before touching anything.

By name or URL

{ "search": "api.example.com", "limit": 50 }

By state (only currently-running monitors)

{ "filter": ["UP", "DOWN", "NOT_STARTED"], "limit": 50 }

Excluding PAUSED avoids re-pausing already-paused monitors.

Follow the instructions field for pagination until hasMore: false. Do not start pausing until the full list is collected — partial pauses mid-paginate are easy to miss on resume.

Step 2 — Confirm with the user

Before any write, echo the list back:

Pausing 14 monitors tagged `deploy-api`:
- Prod API (HTTP, UP)
- Prod API /healthz (KEYWORD, UP)
- ...

Wait for confirmation if the count is non-trivial (>5).

Step 3 — Pause each monitor

{ "monitorId": 800123456, "status": "PAUSED" }

Loop over the collected IDs. If you hit HTTP 429, back off with jitter (see errors skill).

Step 4 — Track what you paused

Before pausing, store the list of monitorIds somewhere retrievable (chat context, a note file, a gist) — you'll need it to resume. Don't rely on re-filtering later: the user may have added or renamed monitors in the interim, and a tag-based filter can quietly pick up unrelated monitors on resume.

Step 5 — Resume

Same loop, flip the status:

{ "monitorId": 800123456, "status": "STARTED" }

Resuming a monitor can return -28001 monitor_limit_exceeded if the account is now over its active-monitor cap (e.g. the user added new monitors while these were paused). Surface which ones failed and suggest deleting or upgrading — do not retry.

Full example flow

User: "Pause everything tagged deploy-web for the next 20 minutes."

1. list-monitors with { "limit": 50 }, paginate, filter client-side where tagNames includes "deploy-web" → 8 monitors.
2. Echo the 8 names and types. User confirms.
3. For each: update-monitor-status with {"monitorId": ..., "status": "PAUSED"}.
4. Note the 8 IDs. Tell the user: "Paused 8 monitors. Run resume deploy-web when you're done."
5. On resume: loop with STARTED. Report which recovered and which errored.

Common mistakes

• Re-filtering on resume instead of using the stored ID list — will silently miss monitors whose tag or name changed.
• Skipping the confirmation step for large batches. Pausing production monitoring without echoing the list is high-risk.
• Pausing monitors already in PAUSED state — harmless but noisy. Filter them out up front.
• Retrying -28001 monitor_limit_exceeded on resume. It's a cap, not a rate limit.
• Using update-monitor with status: ... instead of update-monitor-status. Status lives on the dedicated tool.

Related

• manage-monitors — for the underlying list-monitors and update-monitor-status tools.
• maintenance-windows — the scheduled/recurring alternative.
• errors — for -28001, -31002, and 429 handling.
• incident-response — when pausing is a response to ongoing incidents rather than a deploy.