Model lifecycle observatory
Know when a model dies — before it dies on you.
Your model deprecation radar — it watches the AI models your product depends on and warns you before a shutdown breaks production.
- Get alerted early. Deprecation signals land in your inbox on Free — webhook + email on Indie — so a shutdown shows up as a task, not a postmortem.
- Prove the replacement. One-click migration evals run your real prompts against the new model and diff the outputs — switch with evidence, not hope.
- See your whole exposure. X-Ray finds every pinned model ID in your code and maps it to days-until-shutdown, with a copy-paste migration patch.
Monitoring 58 models across OpenAI, Anthropic, Google, Meta, Mistral, xAI and DeepSeek — with reviewed lifecycle status and provider sources.
No credit card. Free plan: 3 tracked models · 1 migration eval/month · public shutdown calendar. Indie $19: unlimited models, X-Ray repo scan, webhook + email alerts. AI-assisted monitoring, human-verified.
ModelMinder tracks provider deprecation signals for the AI models your product depends on — OpenAI, Anthropic, Google, Meta, Mistral, xAI, DeepSeek, plus video models like Veo, Kling, Seedance and Runway — and proves replacement models with migration evals before you migrate. Free plan available; Indie $19/month; Team $49/month.
What is ModelMinder?
ModelMinder is an AI model deprecation radar. It tracks the AI models your product depends on — from OpenAI, Anthropic, Google, and others — and alerts you before deprecations, renames, and shutdowns break production, with migration evals that prove replacements work before you switch.
Register each model your product depends on and ModelMinder watches provider deprecation signals for you. When a tracked model heads for retirement, you get an inbox alert — webhook and email on paid plans — plus a one-click migration eval that runs your real prompts against the replacement and diffs the outputs, so you switch with evidence instead of hope.
The instrumentsObservation log
Deprecation radar ALL PLANS
A curated catalog of widely-used models with lifecycle status — active, legacy, deprecated, retired — linked to each provider's official changelog so you can verify.
TIER — every plan, including Free. The catalog is the observatory's foundation.
Sunset Risk Score ALL PLANS
A transparent 0–100 deprecation-risk score on every model — status, shutdown proximity, supersession, snapshot age — with the "why" spelled out. Classified like stars: O critical · B high · A medium · F low.
TIER — free on all plans. Scores you see here are the reason to scan your code next.
Codebase X-Ray FREE HOOK REPO SCAN · INDIE+
Paste code or a config and X-Ray finds every pinned model ID, matched against the catalog — file, line number, days until shutdown, suggested replacement, and a copy-paste migration patch.
TIER — paste-scan free forever; whole-repo GitHub scanning is Indie ($19/mo) and up.
Shutdown Calendar PUBLIC FEED · FREE PERSONAL FEED · INDIE+
Every shutdown date as a calendar event with 30/7/1-day reminders. The public catalog feed and RSS are free; your personal feed carries only the models you track.
TIER — public ICS + RSS free; personal signed feed is Indie ($19/mo) and up.
Migration eval 1/MO FREE
One click: your saved prompts run against the replacement model. Deterministic diffs, pass/attention/fail verdicts, a judge summary — proof the replacement works before you switch. Signed read-only report links for your team.
TIER — 1 eval/month free; 10/month on Indie; unlimited* on Team.
Dependency graph + blast radius ALL PLANS
Register the models your product uses and the prompts that use them. When a model is deprecated, ModelMinder counts how many of YOUR prompts are at risk — per model, per registry. In-app alerts on Free; webhook + email on paid plans.
TIER — graph + in-app alerts free; webhook + email alerts are Indie ($19/mo) and up.
Observation methodHow it works
Epoch 1
Register your dependencies
Create a registry for your product, add the models it depends on from the catalog, and save the prompts that use them.
Epoch 2
Get alerted early
When a tracked model moves toward deprecation, you get an inbox alert — webhook and email on paid plans — before the cutoff lands in your postmortem.
Epoch 3
Prove the replacement
Connect the replacement model through any OpenAI-compatible endpoint and run a migration eval on your real prompts. Or X-Ray your whole repo first.
Epoch 4
Switch with evidence
Diffs, similarity scores, and a judge summary show exactly what changes — so migration day is boring, the way it should be.
EphemerisPricing, in one transit
Free
$0 forever
- 3 tracked models · 1 eval/month
- X-Ray paste-scan · Sunset Risk Scores
- Public shutdown calendar + RSS
- In-app deprecation alerts
Indie
$19 /month
- Unlimited tracked models · 10 evals/month
- X-Ray repo-URL scan + migration patches
- Personal calendar feed (your models only)
- Webhook + email alerts · LLM-judge option
Team
$49 /month
- Everything in Indie
- API access · 5 team seats
- Catalog refresh + manual import
- —
Field guidesLearn the deprecation game
Calendar
AI model retirement calendar
Every announced shutdown date in one table — OpenAI, Anthropic, Google — each row linked to the vendor's official source with a last-verified date.
Guides
Deprecation & migration guides
Practical writing: what happens when a model is deprecated, the migration playbook, building a model dependency inventory, and reading sunset-risk signals.
Compare
ModelMinder vs AI gateways
Gateways keep you alive when a model dies. ModelMinder makes sure you knew 60 days ahead and already migrated — an honest comparison.
FAQ
How do I track AI model deprecations across providers?
Register each model your product depends on in ModelMinder. It watches provider changelogs and deprecation notices and alerts you before a cutoff — then runs migration evals so you can prove the replacement works on your own prompts before switching.
Where do deprecation statuses come from?
From a manually curated starter catalog. Every status carries a review date and a link to the provider's official changelog — we don't scrape or claim live provider data. On Team you can refresh and import your own curation.
How does the migration eval call models?
You connect any OpenAI-compatible /chat/completions endpoint (your provider, your key, stored for your workspace). The eval sends your saved prompts to the replacement, diffs the outputs against your reference outputs deterministically, and verdicts each prompt pass / attention / fail.
Is the judge an LLM?
The default judge summary is a labeled heuristic summary — we never pretend a rule-based summary is a model. On paid plans you can point the eval at your own judge endpoint for real LLM verdicts.
What happens when a model I track is deprecated?
A signal is recorded with your blast radius (how many of your prompts use that model), and you get an inbox alert — plus webhook and email on paid plans.
Can I share an eval report with my team?
Yes — every finished eval has a signed read-only link you can revoke at any time.
What if ModelMinder doesn't earn its keep?
One-click refund, no questions asked — message us and your money is back.
How do I pay?
Message us on WhatsApp or email avilanatanael640@gmail.com for a license key, then redeem it on the billing page. Card payments coming soon.
ModelMinder Indie — $19: unlimited models, X-Ray repo scan, webhook + email alerts.