OutageDeck

OutageDeck MCP server

OutageDeck gives AI coding agents current status, incident timelines, and observed uptime history for cloud and SaaS vendors, based on the official status feeds they publish. The remote server needs no API key for public status checks.

Use it to answer “is it us or them?” before debugging your code.

• Remote MCP endpoint: https://outagedeck.com/api/mcp
• Transport: Streamable HTTP
• Public status tools: free, no account or key
• Official MCP Registry: com.outagedeck/outagedeck-status
• Full documentation: outagedeck.com/developers/mcp
• Installable outage-triage agent: published in GitHub Awesome Copilot
• GitHub Actions: outagedeck/status-check@v1
• Terminal and general CI: brew install outagedeck/tap/outagedeck

Every successful check_my_stack response includes two links: links.liveBoard for the current stack view and links.alerts for the same resolved providers already selected in the account flow. Free email alerts cover up to five providers. Slack, Teams, Discord, and webhook delivery are available from the same setup.

Connect in one line

Antigravity CLI (installs the native plugin with public read-only MCP tools):

agy plugin install https://github.com/outagedeck/mcp

The plugin uses Antigravity's current serverUrl schema, requests no secrets or local execution, and disables the five account-scoped tools. See the Antigravity setup guide for prompts and the current public tool catalog.

Gemini CLI (installs the public read-only MCP tools as an extension):

gemini extensions install https://github.com/outagedeck/mcp

The extension requests no settings or secrets and intentionally exposes only the nine public read-only tools. See the Gemini CLI setup guide for prompts and the current tool catalog.

Codex (installs the MCP server plus the dependency-outage triage skill):

codex plugin marketplace add outagedeck/codex-plugins

Then run /plugins, choose OutageDeck, install the plugin, and start a new session. The public plugin marketplace is validated against the live MCP contract.

Claude Code (installs the same MCP server and triage skill):

claude plugin marketplace add outagedeck/codex-plugins
claude plugin install outagedeck@outagedeck

Portable Agent Skill for Codex, Claude Code, Cursor, GitHub Copilot, Windsurf, Gemini CLI, Cline, and other compatible agents:

npx skills add https://github.com/outagedeck/codex-plugins --skill triage-dependency-outages

The portable skill is listed with a passing Snyk audit on skills.sh. It uses the MCP tools when available and otherwise falls back to the anonymous REST API for current status and incident evidence.

Agent Package Manager (after apm init; deploys to the clients declared in your APM project):

apm install --mcp com.outagedeck/outagedeck-status --transport http --registry https://registry.modelcontextprotocol.io

Claude Code, MCP connection only:

claude mcp add --transport http outagedeck https://outagedeck.com/api/mcp

OpenCode: add the remote server to opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "outagedeck": {
      "type": "remote",
      "url": "https://outagedeck.com/api/mcp"
    }
  }
}

Verify the connection with opencode mcp debug outagedeck, then ask: “Check AWS, GitHub, and Cloudflare for active incidents before changing deployment code. use outagedeck”. See the OpenCode setup guide for the current tool catalog and trust model.

For MCP clients that accept JSON configuration:

{
  "mcpServers": {
    "outagedeck": {
      "url": "https://outagedeck.com/api/mcp"
    }
  }
}

No package, daemon, or environment variable is required.

What an agent can do

Ask your agent:

• “Before debugging this deploy, check AWS, Cloudflare, GitHub, and OpenAI.”
• “Is Claude down, or is this failure in my app?”
• “What major vendor incidents are active right now?”
• “Show GitHub's uptime and incident history for the last 30 days.”
• “Give me a citable timeline for the current Cloudflare incident.”

Public tools:

Tool	Purpose
get_provider_status	Vendor-published status for one provider
check_my_stack	One verdict across up to 12 providers, with a prefilled alert setup link
list_active_incidents	Active incidents across the catalog
get_incident_details	A vendor's complete incident update timeline
get_uptime	Independent 7–90 day uptime history
get_outage_report	Cross-vendor reliability summary
search_providers	Find providers by company or product name
search and fetch	Search and retrieve citable OutageDeck documents

Account tools let subscribers manage their own custom providers and alerts without sharing webhook credentials in a conversation: list_custom_providers, add_custom_provider, update_custom_provider, remove_custom_provider, and watch_provider.

Data and trust model

OutageDeck reads official provider status feeds about every 10 minutes. It does not guess from social reports or claim to replace a provider's own status page. Tool results include source and OutageDeck links so an agent can cite what it reports.

Public tools are read-only. Account-specific tools require an OutageDeck API key, and each tool declares its read-only, idempotent, and destructive hints to MCP clients.

Availability and plans

Provider status, incident history, the public API, RSS, badges, and the public MCP tools are free without an account. Free accounts add email alerts for up to five providers. Paid plans add Slack, Teams, Discord, and webhook alerts, custom providers, and higher API quotas.

• Start free email alerts for up to five providers
• See pricing
• Browse all providers

Registry metadata

server.json mirrors the active record published to the official MCP Registry. This repository documents the hosted service; it does not contain the production server implementation. The MIT license applies to the documentation and registry metadata in this repository.

Questions or security reports: [email protected]