For engineering teams
The LLM provider you depend on will break something. You'll hear about it from us first.
One weekly digest tracking deprecations, pricing changes, and breaking schema changes across OpenAI, Anthropic, Google, xAI, Together, Mistral, Groq, Cloudflare Workers AI, and AWS Bedrock — plus an urgent email the same day a provider announces a hard cutoff. $19/mo, cancel anytime.
First digest lands the Monday after you sign up.
What a real entry looks like
These are real, dated changes from providers this year — the kind of thing that ships as a same-day urgent alert or a line in the weekly digest, not a hypothetical.
Urgent — breaking
xAI retired 8 Grok models with ~9 days' notice
Old model slugs now silently redirect to grok-4.3 at different pricing ($1.25 / $2.50 per 1M tokens) instead of erroring — calls kept working, but cost and output quality shifted underneath you with no exception thrown. What to check: pin your model string to a dated snapshot if the provider offers one, and add a cost assertion to your billing alerts, not just an error-rate one.
Urgent — breaking
OpenAI shut down the Assistants API (Aug 26, 2026)
Traffic on the old endpoint now 404s. What to check: anything still calling /v1/threads or /v1/assistantsneeds to move to the Responses API + Conversations API before the cutover in your account — migration path and the field-mapping differences are in this week's email.
Weekly digest
Together AI and Cloudflare Workers AI model sunsets
Together is deprecating Qwen3.5-397B-A17B with MiniMax-M3 as the suggested successor; Cloudflare Workers AI is retiring Kimi-K2.5 in favor of Kimi-K2.6. Neither is urgent yet — both land in the digest with their exact sunset dates so you can schedule the swap instead of getting surprised by it.
Frequently asked
How is this different from just watching each provider's changelog page?
You'd need to check 8+ separate changelog, deprecation, and status pages every week across OpenAI, Anthropic, Google, xAI, Together, Mistral, Groq, and Cloudflare Workers AI, and most of them don't have an RSS feed worth trusting. This puts all of it in one weekly email, plus a same-day alert when a provider announces a hard cutoff instead of waiting for Monday.
Is this just a changelog scraper?
No — generic page-change monitors (the kind built for marketing teams tracking competitor pricing pages) exist and are cheap, but they send you a raw diff and expect you to figure out if it matters. Every APIDrift entry says what actually breaks, which of your calls are affected, and what to change, not just that a page updated.
What counts as a breaking change you'll catch?
Model retirements and sunset dates, parameter/schema changes (renamed fields, changed defaults, removed endpoints), rate-limit restructuring, and pricing changes on models you're likely already calling. Non-breaking additions (new models, new optional params) go in the weekly digest, not an urgent alert.
Which providers do you track?
OpenAI, Anthropic, Google (Gemini), xAI (Grok), Together AI, Mistral, Groq, Cloudflare Workers AI, and AWS Bedrock at launch. Tell us who else you need after signup — coverage expands based on what subscribers actually run in production.
Can I cancel any time?
Yes, it's a monthly Stripe subscription with no contract.