# Model Status Monitor (EZO Labs) > status.ezo.dev probes frontier LLM APIs around the clock and scores each model against its own trailing 7-day normal, never against other models. Everything below is free, read-only and needs no key. Run by [EZO Labs Inc](https://ezolabs.io/). ## What the board measures - Probes run from Cloudflare's edge against each provider's public API: a pulse of 3 tasks every 10 minutes, and a battery of 8 tasks every 2 hours. - Probes stream, time out at 45 seconds and are never retried. - Graders are deterministic string and structure checks. There is no LLM judge. - Per model: p95 and p50 latency, time to first token, round trip, accuracy (pass rate), output tokens, and cost per 1,000 probes. - EZO Health is one composite per model, where 1.00 is exactly that model's own normal and 0 is broken. - A model is called degraded when it moves off its own baseline on errors, latency, accuracy or token drift. - Prices are read off each vendor's own pricing page and carry the date they were checked. Full method, prompts and thresholds: https://status.ezo.dev/methodology ## HTTP API All routes answer CORS from any origin. JSON unless noted. - [GET /api/status](https://status.ezo.dev/api/status): the whole board, every model's live window, baseline and status. - [GET /api/models](https://status.ezo.dev/api/models): the roster and price table, in USD per million tokens, with tiers, peak hours, source and verified date. - GET /api/models/: one model's card, history, per-task results and incidents. Query: days (1 to 90). - [GET /api/incidents](https://status.ezo.dev/api/incidents): the incident log. Also as RSS at /api/incidents.rss. - [GET /api/series?days=7](https://status.ezo.dev/api/series?days=7): every metric bucketed over time, per model. Query: days (1 to 90), bucket (hour or day). - GET /badge/.svg: an embeddable status badge. Model ids look like provider/model, for example anthropic/claude-opus-5-5. ## MCP - Endpoint: POST https://status.ezo.dev/mcp (streamable HTTP, JSON responses, stateless, rate limited per address) - Add it to Claude Code: claude mcp add --transport http ezo-status https://status.ezo.dev/mcp Tools, all read-only: - list_models: the roster with prices (the /api/models payload). - model_status: one model's live row plus its current price. Args: id. - model_series: one model's history. Args: id, days (1, 7, 30 or 90; default 7). - incidents: the most recent incidents. Args: limit (1 to 100; default 25). - board_summary: the tally, the last run, and the models furthest faster and slower than their own normal.