@pipeworx/statuspage
MCP server that answers service status and uptime questions by reading Atlassian Statuspage feeds from hundreds of vendors, with keyless access and tools for checking status, incidents, and vendor listings.
README
@pipeworx/statuspage
Service status / uptime MCP — "is X down right now?" answered from vendors' own Atlassian Statuspage feeds. Keyless.
Atlassian Statuspage is a de-facto standard: hundreds of vendors expose the same
JSON contract at https://<status-host>/api/v2/status.json. This pack reads that
contract, so it closes a class of question rather than one vendor.
Tools
statuspage_check(vendor | status_host, include_components?)— current status for one vendor: overall indicator, the vendor's own status line, every degraded component, open incidents with the latest update text, and upcoming maintenance where published.statuspage_incidents(vendor | status_host, unresolved_only?, limit?, include_all_updates?)— incident history: title, lifecycle status, impact, start/resolve times, duration, affected components, latest update.statuspage_list_vendors(category?, query?, include_elsewhere?)— the vendors this pack can answer for, with each one's status host, so a name can be confirmed rather than guessed.statuspage_multi_check(vendors[])— up to 20 vendors in one call, returned as a compact up/down table plus a roll-up. Hosts are de-duplicated, concurrency capped at 5.
Coverage
Coverage is the curated vendor map plus any caller-supplied host.
The curated map holds 123 vendors across AI/model providers, cloud and hosting, databases and observability, developer tools, payments and fintech, communications, and business SaaS.
Every host in the map was verified, not guessed. 206 candidate hosts were probed
live; each survivor had to return a parseable Statuspage v2 payload on
/api/v2/status.json, /api/v2/summary.json and /api/v2/incidents.json.
Candidates that failed were dropped rather than shipped on a hunch — status
hostnames are genuinely unguessable (status.anthropic.com redirects to
status.claude.com; status.cloudflare.com does not resolve at all, the real host
is www.cloudflarestatus.com).
For anything outside the map, pass status_host: "status.somevendor.com". Every
response says which of the two it used via the source field
(curated_map | caller_supplied_host).
Vendors that publish status elsewhere
Roughly 43 well-known companies were probed and found to publish status in some
other format — AWS, Google Cloud, Azure, Slack, Stripe, GitLab, Heroku, PagerDuty,
Notion, Databricks and others. Asking for one of those returns
reason: 'vendor_not_statuspage' together with the URL a human should open, rather
than nothing. statuspage_list_vendors({include_elsewhere: true}) lists them.
Statuspage "lite" hosts
17 hosts in the map serve status.json / summary.json / incidents.json but omit
/api/v2/incidents/unresolved.json (OpenAI is one). The pack falls back to filtering
the incident feed for open incidents, and reports which path it used in
incidents_source.
Result semantics
- "All Systems Operational" is a positive result, returned as
found: truewithoperational: trueand an emptyopen_incidentsarray. - A vendor with no incidents and a status page we could not read never look the
same. A fetch failure returns
found: false, reason: 'status_page_unreachable'. - An unrecognised vendor returns
found: false, reason: 'vendor_not_covered'with near matches and a pointer tostatuspage_list_vendors— never a guessed host, never a silently substituted vendor.
Data source
Atlassian Statuspage v2 REST API, served by each vendor on its own status host:
/api/v2/status.json— overall indicator + description/api/v2/summary.json— status + components + active incidents + maintenance/api/v2/incidents.json,/api/v2/incidents/unresolved.json/api/v2/components.json
No API key, no auth, no rate-limit handling in the pack.
Quick Start
Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):
{
"mcpServers": {
"statuspage": {
"url": "https://gateway.pipeworx.io/statuspage/mcp"
}
}
}
Or connect to the full Pipeworx gateway for access to all 1422+ data sources:
{
"mcpServers": {
"pipeworx": {
"url": "https://gateway.pipeworx.io/mcp"
}
}
}
Using with ask_pipeworx
Instead of calling tools directly, you can ask questions in plain English:
ask_pipeworx({ question: "your question about Statuspage data" })
The gateway picks the right tool and fills the arguments automatically.
More
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.