Site Qwality
Hosted remote MCP server for uptime monitoring, incidents, logs, metrics and traces. OAuth or scoped API key, included on every plan including free.
Hosted MCP Server
npx add-mcp 'https://mcp.siteqwality.com/mcp'Installs into Claude Code, Codex, Cursor and more
Documentation
Ask about your services in plain words.
The server speaks Streamable HTTP (MCP 2026-07-28) and is hosted only: nothing to install, no local or self-hosted version. Add the URL to your assistant, sign in, choose what it may do, and it works with your monitors, incidents, status pages, logs, metrics and traces. 26 of the 64 tools only read; 38 can change things.
Monitors and checks (12 tools)
See what is wrong, read the evidence and run everyday actions. The tools work on HTTP, heartbeat (cron) and browser monitors; creating and editing is for HTTP monitors.
Read: list_attention_items, list_monitors, get_monitor, get_recent_checks, get_alert_history, list_check_locations
Change: create_http_monitor, update_http_monitor, pause_monitor, resume_monitor, run_browser_check
Delete (flagged destructive): delete_monitor
Incidents (11 tools)
List and read incidents with their timelines, open one, post public updates or internal notes, acknowledge or resolve it, and publish it to a status page. Resolving a published incident posts a public Resolved update and emails the page's subscribers.
Read: list_incidents, get_incident, list_status_page_incidents
Change: create_incident, post_incident_update, acknowledge_incident, resolve_incident, publish_incident, update_status_page_incident
Flagged destructive: unpublish_incident, delete_status_page_incident
Status pages: pages and components (17 tools)
Create and rename pages, change the layout, add manual components, put monitors and heartbeats on a page, and override what a page shows for one component, with an optional note and end time.
Read: list_status_pages, get_status_page, list_status_page_components, list_status_page_monitors, get_status_page_monitor
Change: create_status_page, rename_status_page, create_status_page_component, update_status_page_component, set_status_page_component_status, clear_status_page_component_status, add_status_page_monitor, update_status_page_monitor
Flagged destructive: delete_status_page, update_status_page_layout (replaces the whole layout), delete_status_page_component, remove_status_page_monitor
Status pages: maintenance, subscribers, branding, domain (17 tools)
Schedule maintenance (confirmed subscribers hear about new and rescheduled windows), manage email subscribers, change branding and serve a page on your own domain.
Read: list_status_page_maintenances, get_status_page_maintenance, list_status_page_subscribers, export_status_page_subscribers, get_status_page_branding, get_status_page_custom_domain
Change: schedule_status_page_maintenance, update_status_page_maintenance, add_status_page_subscriber, import_status_page_subscribers, update_status_page_branding, get_status_page_branding_upload_url, set_status_page_custom_domain
Flagged destructive: delete_status_page_maintenance, remove_status_page_subscriber, delete_status_page_branding_asset, remove_status_page_custom_domain
Logs, metrics and traces (5 tools, read only)
Search logs by text, level and source, list and query metrics over a time window, and search traces and open one trace span by span, so the assistant can answer "what changed around 14:00?" from your own data.
Read: search_logs, list_metric_names, query_metrics, search_traces, get_trace
Everything else, through 2 tools
Notifications, on-call schedules, escalation policies, dashboards, RUM, DNS record monitors and the rest of the documented API have no curated tools. find_api_operations searches the API for the operation you described and call_api calls it, with the scope the operation needs.
Read: find_api_operations
Flagged destructive: call_api, because it can run any operation it is allowed to reach
Sign in, no keys to paste
Approve it once,revoke it any time.
Add the server URL and your assistant sends you to Site Qwality to sign in. You see which assistant is asking and what it wants, switch on read-only or untick any area you would rather not grant, and approve.
- The assistant acts as you, and never with more access than your role allows
- A viewer's assistant is as read-only as the viewer, even if write access is approved
- Revoke it under AI assistants in your personal settings; it loses access within a minute
app.siteqwality.com / oauth / authorize
Approve access WaitingLIVE
VIEW 3 areas
CHANGE 1 area
EXPIRES on revoke
monitorsview
incidentsview
pagesview
monitorschange
Setup by client
Paste one URL.
Every client gets the same server URL, https://mcp.siteqwality.com/mcp. Clients that support OAuth open a browser to sign you in the first time. The setup guide has the full steps.
- Claude (web and Desktop): Customize, Connectors, Add custom connector, paste the URL, then Connect. Free Claude plans allow one custom connector
- ChatGPT: add the URL as a custom MCP app with OAuth authentication. Which ChatGPT plans offer custom MCP apps is OpenAI's call and changes, so follow the setup guide for the current steps
- Claude Code, Cursor, VS Code and Codex CLI: use the snippets on this page
Claude Codeclaude mcp add --transport http siteqwality \
https://mcp.siteqwality.com/mcp
# then run /mcp in Claude Code and sign in
Cursor: ~/.cursor/mcp.json{
"mcpServers": {
"siteqwality": {
"url": "https://mcp.siteqwality.com/mcp"
}
}
}
VS Code:.vscode/mcp.json{
"servers": {
"siteqwality": {
"type": "http",
"url": "https://mcp.siteqwality.com/mcp"
}
}
}
Codex CLIcodex mcp add siteqwality \
--url https://mcp.siteqwality.com/mcp
codex mcp login siteqwality
Headless agents and CI
Or use anAPI key.
Agents that cannot open a browser use an API key instead, sent as a Bearer token. Create it under Settings, API keys with Full access (ingest keys cannot call the API). The key's scopes apply to every tool, exactly as they do on the REST API.
- Works with any client that speaks Streamable HTTP and can send a header
- Tick only the scopes the agent needs: a key with no scopes ticked can do everything your account can
- Its changes are audited as that key, marked as coming through MCP
Claude Code$ claude mcp add --transport http siteqwality \
https://mcp.siteqwality.com/mcp \
--header "Authorization: Bearer $SITEQWALITY_API_KEY"
Permissions
Read or change,area by area.
Access is granted as read:<area> and write:<area>. Write does not include read, so an agent that needs both needs both. The consent screen offers the first nine areas; API keys can also take the last two.
- A tool the granted scopes do not cover is refused with a 403, the same as on the REST API
- Adding, changing or removing an integration needs a separate permission that the consent screen never offers
- Session replay content and public dashboard share links are never available to an assistant
| Area | Covers |
|---|---|
| Monitors | HTTP, SSL/TLS, domain expiry, cron and browser checks |
| Incidents | Incidents (including those on status pages), escalation policies, on-call schedules, maintenance windows |
| Notifications | Notification channels and groups, and reading integrations |
| Status pages | Pages with their monitors, layout, maintenance, subscribers, branding and custom domain |
| Logs | Log search, parsers, saved views |
| Metrics | Metric queries and saved views |
| Traces | Trace search and detail |
| RUM | RUM applications and analytics |
| Dashboards | Dashboards, templates, version history, widget queries |
| Services (API key only) | Services and the escalation policy each one routes to |
| User notification settings (API key only) | Each user's contact methods and notification rules |
Security
Every changeis on the record.
An assistant's changes go through the same API as the dashboard, so they land in your audit log under Settings, Audit log. Super admins, admins and auditors can read it, back 90 days.
- Billing, inviting or removing team members, API keys and API clients, SSO and SCIM, account security, closing the account and the staff and sign-in endpoints are withheld from every tool, including
call_api - The 12 tools that delete, remove or replace something are flagged destructive, so assistants that honor the flag ask first. That prompt is the client's behavior, not something we enforce
- Monitor names and log lines come from your own systems. The tools tell the assistant to treat them as data, but that is an instruction to a model, not a guarantee, so keep write access narrow on assistants that read untrusted text
- The log records changes, not reads, and audit writes are best effort: it is a strong record, not a guaranteed-complete ledger
| Audit entry field | What it holds for an assistant |
|---|---|
actor_user_email | The member who approved the assistant |
actor_type | connected_app for an OAuth assistant. An API key used through MCP stays api_key |
connected_app | The assistant's id and the name it registered, for example Claude |
via | mcp |
mcp_tool | The tool that made the change, for example pause_monitor |
oauth.authorize entries | Who allowed or refused an assistant, and the scopes granted |
Sources
Where eachclaim comes from.
Tool names, counts, scopes and sign-in behavior were checked against the MCP server source, the consent screen and the published docs on October 17, 2026. Where a vendor controls the answer, such as which ChatGPT plans offer custom MCP apps, this page says so instead of repeating it.
- Connect AI assistants (MCP), Site Qwality docs: setup, permissions and the full tool list
- Customer audit log, Site Qwality docs: what is recorded for assistants
- Generate API keys, Site Qwality docs: keys and scopes
- API reference, Site Qwality docs: every operation
call_apican reach - Monitoring tools with an MCP server (2026), how other vendors’ servers compare
- MCP specification: transports and authorization, modelcontextprotocol.io
- Client setup: Claude, Claude Code, Cursor, VS Code, Codex CLI
Last verified October 17, 2026.
64
tools for monitors, incidents, status pages and telemetry, 26 of them read only
$0
MCP access on every plan, including free
9
areas the consent screen lets you grant read or write on, one by one
1 min
for a revoked assistant to lose access
FAQ
Yes. It works with any client that supports remote MCP servers over Streamable HTTP, with OAuth sign-in or an API key. The setup guide has steps for Claude (web and Desktop), Claude Code, ChatGPT, Cursor, VS Code and the Codex CLI. In Claude you add the URL under Customize, Connectors, Add custom connector; in Cursor you add it to mcp.json. ChatGPT needs a custom MCP app, and which ChatGPT plans offer one is up to OpenAI, so check the guide for the current steps.
Only if you allow it, and only what your own role allows. At sign-in you can switch on read-only or untick individual areas. 26 of the 64 tools only read. The other 38 can change things, for example pausing a monitor, acknowledging or resolving an incident, posting a status page update or scheduling maintenance. Billing, team membership, API keys, SSO and session replay content are withheld from the tools.
Yes. It is included on every plan, including the free tier, with no separate charge. Plan limits and usage quotas apply exactly as they do in the dashboard. Your AI assistant's own subscription is separate and is not ours.
In the dashboard under Settings, Audit log. Super admins, admins and auditors can read it, for the last 90 days. Every change an assistant makes (any POST, PUT, PATCH or DELETE) is recorded as the member who approved it, with the assistant's name, via: mcp and the tool it used, and connecting or disconnecting an assistant is recorded too. Filter on actor_type=connected_app to see what assistants did. Reads are not logged, so the log shows what an assistant changed, not what it looked at. See the audit log docs.
Whatever the tools return for your account, limited to the areas you granted and to your role: monitor names, URLs and check results, incidents with their timelines, status pages, and logs, metrics and traces if you grant those areas. A single answer is cut to 100 KB, and the answer says when lists were shortened. It cannot see session replay content, billing, team members or API keys. What the assistant does with the data afterwards is governed by the AI provider you chose, not by Site Qwality.
No. It is a hosted server at https://mcp.siteqwality.com/mcp. There is no stdio package and no self-hosted version.
With an API key sent as Authorization: Bearer <key>. Create it under Settings, API keys, choose Full access, and tick only the scopes the agent needs. A key with no scopes ticked can do everything your account can. Its changes are audited as that key and marked as coming through MCP.
The MCP server adds no limit of its own. Every tool call is an ordinary API request made as you, so plan limits, usage quotas and the API's rate limiting apply. We do not publish a per-minute figure.
Several monitoring vendors publish one. Ours has 64 tools, is hosted only and is free on every plan. Others have more tools, a local server option, or port and TCP checks that we do not have. The comparison of monitoring tools with an MCP server lists each vendor from its own docs, dated October 17, 2026.
Revoke it on the AI assistants tab in your personal settings. It cannot sign in again without your approval, and the access it already had stops working within about a minute.