Phase 1 — Watch — Live in Phase 1
ModelSwap Sentinel polls OpenAI, Anthropic, and Google Gemini deprecation and changelog sources around the clock. Every change is diffed against the last snapshot and recorded as a timestamped drift event — so you hear about a retirement when it is announced, not when your production calls start failing.
Three layers, built in order. Only the first one is live today — and we label the other two as exactly what they are.
A cron job polls provider changelog and deprecation pages, snapshots what each page says, and diffs every new snapshot against the last. Any difference becomes a drift event with an evidence excerpt, a timestamp, and the source it came from — delivered to subscriber webhooks (Slack-compatible) the moment they are detected.
When a successor model appears, Prove will run your integration against the old and new models and report behavior regressions side by side — on your prompts and your test cases, not generic benchmarks. This layer is in development and is not live yet.
Patch will generate a migration pull request — model strings, parameters, and breaking-change fixes — backed by the Prove evidence, plus a migration-safe certification when the change passes. This layer is in development and is not live yet.
Real data from the watcher API — the providers under watch right now, and the most recent drift events it has recorded. Nothing on this section is hand-written.
Loading providers…
Loading drift events…
Watch is what you can buy and use today. The Production tier includes Prove and Patch as they launch — it does not pretend they have launched already.
$99/mo
per integration
$499/mo
per integration
Custom
per-migration success fees