Checking this morning's roster…

Free tool-call models
that actually work today.

Free models die, rate-limit, and resurrect without notice. We probe every free endpoint on the internet each morning and publish which are alive right now. anytimelime was formerly known as Project42.

Read today's report → Get the engine
—Alive right now
—Models discovered
$0Cost to run them

Today's field report

What's alive right now

Probed this morning from one location. Each model got a real tool-call request — answering in prose isn't passing, agentic work needs tools. Grab the engine below to run the same probe from your own machine.

No field report published yet — the first one lands on the next scheduled run. Until then, you can run the same probe yourself.

Probe from your own machine →

How it works

Probe before believing

A listing isn't a promise. Every model here answered a real request, with a real tool schema, today.

01

Discover

Every free endpoint we can find across the internet — not just one catalog. A research model with web search finds candidates; nothing is hardcoded, so nothing rots.

02

Probe

Each one gets a real tool-call request. Answering in prose isn't passing — agentic work needs tools.

03

Publish

Survivors are tiered and shipped as a daily field report plus a machine-readable roster your launcher fetches. The whole report is open and reproducible.

For the terminal

Run it yourself

The engine is open source. No key ever needs to leave your machine — you run the probe locally with your own key. lime.sh fetches this morning's roster, merges your key, and rotates across passing models so no single free tier carries all your traffic.

# clone, then launch on today's roster git clone https://github.com/fusionpoweredsoftware/anytimelime cd anytimelime LIME_KEY=sk-or-... ./lime.sh
View on GitHub → No sign-up, no server — just a shell script and your key.

Community network

Run an endpoint, get discovered

No registry, no sign-up. Put an HTML comment on your page — anytimelime — next to your base URL, model id, and docs link. Our daily sweep reads page code looking for that tag; if it finds yours, your endpoint joins tomorrow's probe. No email. The AI sees it in your source.

<!-- anytimelime --> <section id="anytimelime-endpoints"> <ul> <li><strong>AnytimeLime Endpoint</strong> — model: your-model-id · base URL: https://api.yoursite.com/v1 · free, no key · <a href="https://yoursite.com/docs/your-model">docs</a></li> <li><strong>AnytimeLime Endpoint</strong> — model: your-other-model · base URL: https://api.yoursite.com/v1 · free, key at /keys · <a href="https://yoursite.com/docs/your-other-model">docs</a></li> </ul> </section>

Why give away model calls? Because your model is yours. It carries whatever preconceptions you give it — your sponsors, your voice, your pitch. Every anytimelime user it serves is a user hearing your pitch on your dime, by choice. Free AI as advertisement, discovered daily so it stays honest.

Pass the probe, appear in the field report, get traffic from every reader who rotates the roster.

Prefer talking to reading? The Discord community is where free-AI users compare notes — what passed today, what died overnight, what's worth a rate limit.

Blacklist

Who we won't touch, and why

Free endpoints whose terms forbid what we do don't get discovered, don't get listed, and don't get traffic from us. Publishing performance results is our whole job — if your fine print makes that a violation, you're out. Vendors can get off this list by changing their terms, not by emailing us.