List monitoring checks for the organization, optionally filtered by name/slug substring, labels, or check group. Returns {data, pagination}; pass pagination.cursor back as cursor for the next page. Use this for browsing or filtering a fleet of checks. To investigate a single check's current health, use diagnose_check instead; if you already know its UID or slug, get_check is more direct. Read-only: works with mcp:read tokens.

Get a single check's metadata by UID or slug; ask for extras with with=lastResult,lastStatusChange. For a full triage briefing (current status + recent results + active incidents), prefer diagnose_check; to search by name or label, use list_checks. Read-only: works with mcp:read tokens.

Create a new monitoring check and return it (uid, slug, placement, resolved periods). If you don't know what config shape a given type expects, call list_check_types with type first to fetch a working starting config, then use validate_check to dry-run before creating. An enabled check starts running on its next tick and opens an incident after its confirmation period; pass enabled:false to create it paused. A duplicate slug is rejected with a conflict, never merged — use update_check to change an existing check. Requires the mcp scope (mcp:read tokens are refused) and at least the user role in the organization.

Update an existing check by UID or slug and return the updated check. PATCH semantics — only the fields you pass are modified, others stay as-is. Use create_check to make a new check and delete_check to remove one; to pause without deleting, set enabled:false instead. Requires the mcp scope (mcp:read tokens are refused) and at least the user role in the organization.

Soft-delete a monitoring check by UID or slug. The check stops running immediately, its active incidents are resolved and its scheduling jobs removed; historical results are kept. A check that other checks tunnel through is refused. To stop a check without losing it, use update_check with enabled:false instead. Returns {deleted: true, identifier}. Requires the mcp scope (mcp:read tokens are refused) and at least the user role in the organization.

Dry-run validate a check config without creating or changing anything — a pure read, no side effects. Returns {valid, fields, warnings}: valid is false exactly when fields lists the blocking findings (every one of them, not just the first), while warnings are advisory notes that never block. Use this before create_check when you've assembled a config from scratch or modified a sample, to catch problems early. Read-only: works with mcp:read tokens.

Run a check once, now, in every region, instead of waiting for its next period (use it right after a fix, or on a freshly created check). The run goes through the normal result path: it can open or resolve an incident like any scheduled run. Rate limited per check and per organization. Returns {requestedAt, regions: [{region, status}]} where status is queued (due now, or right after the run in progress); read the outcome with diagnose_check or list_results (a result with periodStart >= requestedAt). Requires the mcp scope (mcp:read tokens are refused) and at least the user role in the organization.

Return everything an operator would want to look at to diagnose a single check's current state in one call: current status, recent raw results across regions, any active incident, and the most recent resolved incident. Use this instead of chaining list_results + list_incidents when a human asks "what's wrong with check X?". Returns {check, freshness, regionalIssue, recentResults, activeIncident, lastResolvedIncident} — freshness and regionalIssue are omitted when there is nothing to report, the incidents are null when there are none. Read-only: works with mcp:read tokens.

Query monitoring results (raw or aggregated) with filtering by check, type, status, region, period type, and time range. Use this for trend analysis or to inspect a specific window. For investigating a single check's current state, use diagnose_check instead. Returns {data, pagination, window, effectiveFilter}: data holds the result rows, window the time bounds actually queried, and effectiveFilter the period filter that ran after defaulting. Defaults: if periodType is omitted it falls back to "hour"; if periodStartAfter is omitted it falls back to a window matched to the finest periodType requested (raw=1h, hour=24h, day=30d, month=365d). Read-only: works with mcp:read tokens.

List internal incidents (active or resolved) for the organization, filtered by check, state, or time range. Returns {data, pagination} — total, size, and a cursor naming the last row of a full page; the cursor is not read back on the next call yet, so move forward by narrowing since/until. Use diagnose_check for one check's incidents plus current health in one call, and get_status_page include=incidents for the customer-facing entries on a status page. Read-only: works with mcp:read tokens.

Get a single incident by UID; an unknown UID fails with an "incident not found" error. Returns the incident object — number, state, timestamps, failure counts — plus the check with with=check and up to 50 timeline events with with=events. include=["notifications"] adds the incident's notification attempts (one row per dispatch target with its delivery status, channel and recipient, newest first), filtered by status and capped by limit. Use list_incidents to search and filter across incidents, and diagnose_check for a check's full triage briefing. Read-only: works with mcp:read tokens.

List all monitoring check types supported by this server (e.g. http, tcp, dns, icmp, ssl), or with type, the ready-made sample configs of that type. Call it without arguments first when you don't know what type to use: it returns {data: [...]}, one entry per type with its description, enabled state, period bounds and secretFields. Then call it with type=: it returns {data: [...]} with one entry per matching type carrying its samples, each a complete, valid config to clone and modify (much more reliable than guessing field names). Read-only: works with mcp:read tokens.

List monitoring regions available to the organization (e.g. eu-west-1, us-east-1). Returns {data: [...]} — each region's slug, label and per-region metadata — plus defaultRegions, the org's default placement set. Use these slugs in the regions array of create_check or update_check. Read-only: works with mcp:read tokens.

List all check groups for the organization. Check groups bundle related checks together for shared incident handling and dashboard organization, and each entry carries its slug, rolled-up status and check count. Returns {data: [...]}. Read-only: works with mcp:read tokens.

Explore a target and test a js check before create_check, without saving anything. kind=script runs a js check script exactly as a SolidPing worker would and returns status, output (with console) and metrics; env carries plaintext parameters and secrets only NAMES the secret parameters, which run empty. kind=fetch GETs a URL from the server. kind=browser snapshots a page to pick selectors for browser steps. Reaches the URL or script target you name. Read-only: works with mcp:read tokens.

List every status page in the organization, newest first, and return {data: [...]}. Use get_status_page for a single page by UID or slug (with=sections or include=sections for its contents). Read-only: works with mcp:read tokens.

Get one status page by UID or slug and return it. with=sections embeds its sections and their resources. include adds lists next to the page fields: "sections" (section metadata, display order), "resources" (the resources of the section named by sectionIdentifier) and "incidents" (the customer-facing incident publications on the page, filtered by state and active). Use list_status_pages to browse every page, and manage_status_page_section / _resource / _incident to change them. Read-only: works with mcp:read tokens.

Create a new status page for the organization and return the created page. A duplicate slug is rejected with a conflict, never merged. Only one page can be the default: the first page always becomes it, and isDefault:true demotes the previous default. Every new page is seeded with a default "Services" section and starts enabled and public — use manage_status_page_section to add more sections and update_status_page to edit the page later. Requires the mcp scope (mcp:read tokens are refused) and at least the user role in the organization.

Update an existing status page by UID or slug and return the updated page. PATCH semantics — only fields you pass change, omitted fields keep their current values; the page must already exist and a slug that collides with another page is rejected. Use create_status_page to make a new page and delete_status_page to remove one instead. Requires the mcp scope (mcp:read tokens are refused) and at least the user role in the organization.

Soft-delete a status page by UID or slug: the page disappears from the API and the public site, taking its sections and resources with it, and its uploaded brand assets are deleted. There is no undelete — use update_status_page with enabled:false to take a page down without destroying it. Returns {deleted: true, identifier}. Use delete_status_page_section or delete_status_page_resource to remove part of a page instead. Requires the mcp scope (mcp:read tokens are refused) and at least the user role in the organization.

Create, update or delete a section of a status page. action=create adds a section (name and slug required) and returns it. action=update edits one (sectionIdentifier required; omitted fields are kept) and returns it. action=delete permanently removes a section and every resource pinned in it, returning {deleted, identifier}. Read a page's sections with get_status_page include=sections. Annotated destructive and non-idempotent because delete is one of its actions. Every action requires the mcp scope (mcp:read tokens are refused) and at least the user role in the organization.

Pin, edit or unpin a resource (a check or a check group) in a status page section. action=create pins exactly one of checkUid or checkGroupUid and returns the resource. action=update edits one (resourceUid required; omitted fields are kept). action=delete unpins it, returning {deleted, identifier}; the check or group itself is untouched. Read a section's resources with get_status_page include=resources. Annotated destructive and non-idempotent because delete is one of its actions. Every action requires the mcp scope (mcp:read tokens are refused) and at least the user role in the organization.

Manage the customer-facing incidents a status page shows (publications), distinct from the internal incidents list_incidents returns. Titles and bodies are shown to CUSTOMERS: never paste probe output, error strings, internal hostnames or IPs. action=create publishes a hand-written incident (pageIdentifier, title). action=update edits a publication's title, severity or state (pageIdentifier, publicationUid; PATCH semantics). action=add_update appends a narrative entry (pageIdentifier, publicationUid, kind, bodyMarkdown); entries are append-only. action=publish relays an EXISTING internal incident (incidentUid, pageIdentifier) with a title templated from the page. action=unpublish removes a relayed incident from the page (incidentUid, publicationUid), keeping the row for audit. create, add_update and publish notify status-page subscribers (hourly cap) and every action reaches webhook connections. Read publications with get_status_page include=incidents. Annotated destructive and non-idempotent because delete is one of its actions. Every action requires the mcp scope (mcp:read tokens are refused) and at least the user role in the organization.