# Prismix Status MCP **Ask Claude "Is OpenAI down?" and get a live answer.** [![Listed on the official MCP Registry](https://img.shields.io/badge/MCP%20Registry-io.github.Prismor%2Fprismix--status--mcp-2ea44f)](https://registry.modelcontextprotocol.io) [![Glama](https://glama.ai/mcp/servers/Prismor/prismix-status-mcp/badge)](https://glama.ai/mcp/servers/Prismor/prismix-status-mcp) A Model Context Protocol (MCP) server exposing real-time operational status for **75+ AI services** — OpenAI, Anthropic, Cursor, GitHub Copilot, Gemini, Mistral, Perplexity, DeepSeek, Hugging Face, and more. Powered by [Prismix](https://prismix.dev), which monitors each provider's official status feed around the clock. - **Hosted endpoint, zero setup:** `https://prismix.dev/api/v1/mcp` - No API key, no signup, no rate-limit hoops - Streamable HTTP transport, MCP protocol `2024-11-05` - CORS-enabled; works from Claude Desktop, Claude Code, Cursor, or any MCP client - This repo is the server's source — also deployable as your own Cloudflare Worker ## Quick start ### Claude Desktop Add to `claude_desktop_config.json` (Settings → Developer → Edit Config): ```json { "mcpServers": { "prismix-status": { "url": "https://prismix.dev/api/v1/mcp" } } } ``` Restart Claude Desktop. Done — ask *"Which AI services are degraded right now?"* ### Claude Code ```bash claude mcp add --transport http prismix-status https://prismix.dev/api/v1/mcp ``` ### Cursor Add to `.cursor/mcp.json` in your project (or `~/.cursor/mcp.json` globally): ```json { "mcpServers": { "prismix-status": { "url": "https://prismix.dev/api/v1/mcp" } } } ``` ## Tools ### `check_ai_status` Current status for one service (or all of them). | Argument | Type | Description | |---|---|---| | `service` | `string` (optional) | Service ID or partial name, case-insensitive — `"openai"`, `"anthropic"`, `"cursor"`. Omit to check everything. | Returns per service: status indicator (Operational / Degraded / Partial outage / Major outage), active incident count, 30-day uptime %, response latency, and a link to the full incident history. **Things you can ask:** - *"Is Anthropic down right now?"* - *"What's Cursor's 30-day uptime?"* - *"Check the status of every AI service I might be using"* ### `get_incident_history` Recent incident history (last 30 days) for one service — past outages and degradations with impact level, start time, and resolution/duration, plus the 30-day uptime %. | Argument | Type | Description | |---|---|---| | `service` | `string` (required) | Service ID or partial name — `"anthropic"`, `"openai"`, `"cursor"`. | **Things you can ask:** - *"Has Anthropic been reliable this month?"* - *"How often does Cursor go down?"* - *"Show me OpenAI's recent incidents"* ### `list_ai_services` Enumerates all 75+ tracked services with ID, name, category, and current indicator. Useful for discovering exact service IDs. ## Protocol details - JSON-RPC 2.0 over HTTP POST (Streamable HTTP transport) - Protocol version `2024-11-05` - Supported methods: `initialize`, `ping`, `tools/list`, `tools/call` - Notifications (requests without `id`) are accepted with `202` Try it from the command line: ```bash curl -s https://prismix.dev/api/v1/mcp \ -H "Content-Type: application/json" \ -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"check_ai_status","arguments":{"service":"openai"}}}' ``` ## Deploy your own The worker sources data from Prismix's free public REST API (`GET https://prismix.dev/api/v1/statuses` — no auth, CORS-enabled), so a self-hosted copy needs zero secrets: ```bash npm install npx wrangler deploy ``` Your worker URL then speaks the same MCP protocol as the hosted endpoint. ## Data Status data refreshes every ~5 minutes from each provider's official status page or a direct HTTP probe where no status page exists. The full dashboard, incident history, per-service uptime, and embeddable badges live at [prismix.dev](https://prismix.dev). Email alerts on status changes are free with an account. ## License [MIT](./LICENSE)