Monitoring data for your team's sites (status, uptime, incidents, daily insights and domains) and the writes a team makes, for your agents and scripts. A key sees only its own team. Competitors are never included.
A public site's status page numbers need no key: /status/<slug>/data.json and data.csv carry its daily uptime, response time and outages for 90 days (CC BY 4.0).
Authentication
Create a key on the team dashboard's API keys page (team sign-in needed). It's shown once. Send it on every request:
Authorization: Bearer sm_…
A Viewer key reads (GET and HEAD). On teams with 10 sites or more, a Member key can also dismiss incidents, ask for a check again and start or end maintenance windows, and an Admin key can also pause and resume sites: the POST routes below. A key never does more than its creator can now. A signed-in team member's browser can call the GET routes with its cookie; writes need a key. This page, /api and /api/openapi.json need no key.
Conventions
A success is {"data": …}. Lists add limit, offset and has_more; limit is clamped to 1–100.
An error is RFC 9457 problem details (application/problem+json): {"type", "title", "status", "detail", "error": "code", "message": "…", "suggestion"?: "…"}. Branch on error; show message (the same as detail).
Successes and every error carry next_steps: [{"method", "path", "why", "tool"?}], the calls that usually come next or fix the error; tool names the MCP tool that makes the call.
Requests a minute per key, by the team's sites: 60 on the free site, 300 from 2 sites, 1,000 from 25; then 429 with Retry-After; writes are also capped at 10 in 10 seconds per key. Every keyed answer carries RateLimit-Policy and RateLimit (IETF draft 11): "reads";r=57;t=23 means 57 left, the window ends in 23 seconds. The data changes once a minute, so poll at most once a minute.
The API is unversioned and changes only by adding routes and fields; one that has to go is announced at least 90 days ahead, on its responses, with Deprecation (RFC 9745) and Sunset (RFC 8594) headers and a Link to /developers#versioning (rel="deprecation").
Writes are idempotent: repeating one changes nothing more, so retrying after a timeout is safe.
Times are ISO 8601 in UTC (…Z). Durations are milliseconds, sizes bytes. Sites are named by their slug.
Incidents the team dismissed as noise are included, marked dismissed: {by, at}.
Nothing is cached: responses carry cache-control: no-store. New fields may be added; existing ones keep their meaning.
Errors
error
Status
When
unauthorized
401
No key, or an unknown or revoked one.
forbidden
403
A write by a key whose role or team can't make it (writes need a Member or Admin key and 10 sites), a POST to a route that only reads, a write without a key, or a form post from another site.
not_found
404
An unknown site slug or API path.
method_not_allowed
405
Anything but GET or HEAD on a page.
conflict
409
A write this incident can't take (one that can't be checked again, say).
validation_error
400
A parameter isn't valid; `issues` says which.
request_too_large
413
A request body over 64 KB.
rate_limited
429
More requests this minute than the key's team's sites allow (comes with Retry-After): 60 on the free site, 300 from 2 sites, 1,000 from 25.
auth_unavailable
503
The key or sign-in couldn't be checked right now (comes with Retry-After). Your key or session is probably fine: retry.
MCP
https://sitesinthewoods.com/mcp is a hosted MCP server with one tool per route below (the writes for Member and Admin keys only), named by its operation id (status, sites_by_slug…) and returning the route's own JSON. It takes the same key, sent the same way; listing its tools needs none. It speaks protocol 2026-07-28 and the 2025 revisions. Its card is at /mcp/server-card.
Agent skills teach an agent the jobs these routes do (checking status, connecting, finding what's wrong, holding alerts), each one a SKILL.md it can install. The checks behind this data are open source: agent-ready-uptime.
Routes
GET /api/status: Every one of the team's sites: its status right now, and the open incidents
GET /api/sites: The team's sites: status, homepage uptime over 1, 7 and 30 days, open incidents
GET /api/sites/:slug: One site: its last 7 days (as in the weekly brief), open incidents, domain and TLS expiry
GET /api/sites/:slug/insights: One site's daily insights: page weight, security headers, prefetches, page errors, broken links, cache, dead ends
GET /api/sites/:slug/scans: One site's API recovery scans, newest first: state, how each ended, and findings by weight
GET /api/sites/:slug/scans/:id: One API recovery scan: what it saw, its findings, the fixes and what changed since the last scan
Every one of the team's sites: its status right now, and the open incidents. Call this first, or on a schedule (at most once a minute, as often as the checks run), to see whether anything is down or warning right now.
The team's sites: status, homepage uptime over 1, 7 and 30 days, open incidents. Call this to compare sites or report on uptime over the last day, week or month. It also lists the slugs every other route takes.
One site: its last 7 days (as in the weekly brief), open incidents, domain and TLS expiry. Call this to answer questions about one site: how it did this week, what's wrong with it now, and when its domain or certificate runs out.
One site's daily insights: page weight, security headers, prefetches, page errors, broken links, cache, dead ends. Call this when a site is slow or heavy, or to find what to fix: the daily browser load's page weight, missing security headers, JavaScript errors and broken links.
One site's API recovery scans, newest first: state, how each ended, and findings by weight. Call this to see whether a site has been scanned for dead ends (API errors that don't say how to recover, lists without paging, pages nothing links to) and how the last scan came out. Scans are started from the dashboard by an admin or owner.
One API recovery scan: what it saw, its findings, the fixes and what changed since the last scan. Call this for a scan's results: findings by weight with the places on the site, three to five fixes in words and as a block for a coding agent, and new, fixed and still-open findings since the previous scan. It never includes text from the scanned site.
One site's maintenance windows: on now, scheduled, and the last 30 days. Call this to see whether a site is in a maintenance window (its alerts held, its uptime not counting that time) and when the next one is. Windows are started from the dashboard or a deploy calling the site's maintenance URL (POST <url>/start?minutes=30, POST <url>/end), which the dashboard shows to admins.
One site's structure from its sitemaps: sections, new and removed pages, what fails. Call this to see where a site is changing or breaking without crawling it: every path with pages under it, with new and removed pages in the window, pages that fail, broken links and sitemaps that miss pages. Add ?section= for one section's pages. Our sites only (competitors' sitemaps aren't read).
What changed on one site: incidents opened and resolved, code pushes, maintenance, new pages. Call this to find what changed before a number moved (clicks fell, pages got slower): the site's timeline, newest first, the same events the dashboard draws on its charts. Page through with limit and offset.
Incidents, open now or recent; open outages and slowdowns carry the likely cause. Call this to see what went wrong and why: open and recent incidents, newest first, with the likely cause on open outages and slowdowns. Page through with limit and offset.
The team's domains: registration expiry, registrar, nameservers, status. Call this to check renewals: which domains expire soon, who the registrar is, and whether nameservers or registry locks changed.
Dismiss an open incident as noise. Hides it from the open lists for the whole team; it never alerts again and closes without a message. Not every kind can be dismissed (409). Needs a Member or Admin key, on a team with 10 sites or more.
curl -X POST -H "Authorization: Bearer sm_…" https://sitesinthewoods.com/api/incidents/INCIDENT-ID/dismiss
Parameter
In
Type
Meaning
id
path
string
incident id, as in /api/incidents
Response fields (11)
Field
Type
Meaning
data
object
data.id
string
data.kind
string
data.site
string or null
its site's slug
data.dismissed
boolean
data.recheckRequested
boolean
next_steps
object[] (optional)
calls that usually come next ({method, path, why, tool})
next_steps[].method
string (optional)
next_steps[].path
string (optional)
next_steps[].why
string (optional)
next_steps[].tool
string (optional)
the MCP tool that makes this call, for data routes
Check an incident again now. The monitor checks it again within 30 seconds, after a fix, instead of waiting for the next scheduled check. Not every kind can be checked again (409). Needs a Member or Admin key, on a team with 10 sites or more.
curl -X POST -H "Authorization: Bearer sm_…" https://sitesinthewoods.com/api/incidents/INCIDENT-ID/recheck
Parameter
In
Type
Meaning
id
path
string
incident id, as in /api/incidents
Response fields (11)
Field
Type
Meaning
data
object
data.id
string
data.kind
string
data.site
string or null
its site's slug
data.dismissed
boolean
data.recheckRequested
boolean
next_steps
object[] (optional)
calls that usually come next ({method, path, why, tool})
next_steps[].method
string (optional)
next_steps[].path
string (optional)
next_steps[].why
string (optional)
next_steps[].tool
string (optional)
the MCP tool that makes this call, for data routes
Start a maintenance window on a site now. Call this before a deploy: alerts for the site are held (checks still run) until the window ends, after `minutes` (default 30, at most 24 hours) or when you end it. Starting again replaces the open window. Needs a Member or Admin key, on a team with 10 sites or more.
End a site's maintenance windows now. Call this after a deploy: alerts for the site go out again from now. Ending with no window on answers `ended: 0`. Needs a Member or Admin key, on a team with 10 sites or more.
curl -X POST -H "Authorization: Bearer sm_…" https://sitesinthewoods.com/api/sites/SITE-SLUG/maintenance/end
Parameter
In
Type
Meaning
slug
path
string
site slug, as in /api/sites
Response fields (7)
Field
Type
Meaning
data
object
data.ended
integer
windows ended
next_steps
object[] (optional)
calls that usually come next ({method, path, why, tool})
next_steps[].method
string (optional)
next_steps[].path
string (optional)
next_steps[].why
string (optional)
next_steps[].tool
string (optional)
the MCP tool that makes this call, for data routes
Pause a site's checks. Stops every check of the site, keeping its history, until it's resumed. Nothing alerts while it's paused. Needs an Admin key, on a team with 10 sites or more.
curl -X POST -H "Authorization: Bearer sm_…" https://sitesinthewoods.com/api/sites/SITE-SLUG/pause
Parameter
In
Type
Meaning
slug
path
string
site slug, as in /api/sites
Response fields (8)
Field
Type
Meaning
data
object
data.slug
string
data.paused
boolean
next_steps
object[] (optional)
calls that usually come next ({method, path, why, tool})
next_steps[].method
string (optional)
next_steps[].path
string (optional)
next_steps[].why
string (optional)
next_steps[].tool
string (optional)
the MCP tool that makes this call, for data routes