Free forever · No credit card

Your AI bill
could be $0.

24 providers hand out real free tiers. The Free Token Agent signs up, collects the keys on your machine, and routes 300+ models through one local key.

macOS · Linux — Windows on the way

OpenRouterGroqMistralGoogle GeminiGitHub ModelsCerebrasCohereDeepSeekNVIDIA NIMHugging FaceCloudflareOllama CloudSambaNovaAI21
01 — How It Works

Press the button. Keep the keys.

Signing up for free tiers by hand means twenty tabs, twenty verification emails, and twenty dashboards. FreeTokens does the tab-juggling and hands you the wheel only when a human is genuinely required.

01

Press Get Tokens

The agent opens each provider in the app's own browser and registers with your real name and email — filling forms, reading verification codes, copying the key the moment it appears.

02

Help only when asked

A CAPTCHA, an SMS code, a Google sign-in. The agent pauses, asks you in-app, and resumes the second you're done. 13 of the 24 providers need no help at all.

03

Keys stay local

Every key lands in a vault under ~/.freetokens. No cloud, no account, no telemetry. There is no Lovie server behind the app at all.

04

Use one key everywhere

A single lovie_ key on an OpenAI-compatible endpoint routes each request to the strongest free model, and fails over automatically when a tier runs dry.

02 — Inside The App

Watch it work. Then use it all.

The FreeTokens Activity view: the Free Token Agent narrates each step live and captures a Cohere API key

Live, step by step.

The Activity view narrates every move the agent makes: opening the provider, signing in, reading the verification code, revealing the key. You see the moment each key is captured, and you can step in at any point.

18 keys in the vault · 342 models unlocked

The FreeTokens Chat view: one lovie_ key auto-routed across free providers, answering through Cohere

One place for everything.

Chat streams through your own keys, auto-routed to the strongest free model available. Pin a provider when you want control, or let the gateway pick. Documents, agents, voice, and images live right next to it.

One lovie_ key · every model you collected

03 — The Gateway

One key. Hundreds of free models.

Point any OpenAI-compatible client at your machine and forget provider dashboards exist. Llama, Qwen, Gemini, Mistral, DeepSeek, and more behind one local endpoint.

# any OpenAI-compatible tool
export OPENAI_BASE_URL="http://localhost:8787/api/v1"
export OPENAI_API_KEY="lovie_..."

Works with Cursor, Claude Code, Codex, Aider, Cline, Continue, Zed, the OpenAI SDK, and anything that speaks MCP.

Route + failover

Requests go to the strongest available model. When one free tier rate-limits, the gateway retries on the next.

Live web search

A fast planner spots questions that need fresh information and grounds the answer with live search.

Documents + RAG

Drop in files once. They're chunked, embedded, and OCR'd locally. Ask across everything, with sources.

Shareable agents

Build an assistant with its own persona and documents, then share it with a link. It answers on your keys.

Voice + images

Whisper speech-to-text and FLUX image generation ride the same free tiers, through the same endpoint.

MCP server

A local MCP server exposes the provider directory, key guides, and harvest runs to any MCP-capable tool.

04 — Provider Directory

24 providers. Real free tiers.

Every model in the directory was probed with a real key before it made the list. 13 providers harvest with zero clicks; the rest ask you for exactly one thing.

Mistral AI

About 1 billion free tokens a month.

The Experiment plan is the largest free tier in the directory: 42 verified models, from Codestral to Devstral, collected automatically.

Groq

2,600 tokens a second.

1,000 requests a day on the fastest inference of any free tier.

Google Gemini

1,500 requests a day

With a 2-million-token context window. One Google sign-in.

Cerebras

1M tokens a day

Up to 14,400 requests daily on wafer-scale hardware.

OpenRouter

200 requests a day

Dozens of community free models behind a single key.

GitHub Models

GPT on your GitHub account

GPT and Llama models, 50 requests a day, zero clicks.

NVIDIA NIM

40+ verified models

Nemotron and friends with a 1M-token context. One SMS code.

Cloudflare Workers AI

10,000 Neurons a day

45 models, including image generation. One sign-in.

LLM7 + OVHcloud

No key needed at all

Open endpoints that just answer. OVHcloud is EU-hosted.

And 14 more

Cohere, DeepSeek, Ollama Cloud, Kilo Code, Aion Labs, OpenCode Zen, Hugging Face, SambaNova, AI21, Nscale, Alibaba, Z AI, SiliconFlow, and ModelScope. Limits are set by each provider and can change.

05 — Why It's Free

Free, local, and private. On purpose.

FreeTokens is a free tool from Lovie, the company that forms and runs US companies for founders. We build things builders actually use. No trial clock, no upsell inside, no catch.

100% local

The gateway, the database, and every key live under ~/.freetokens on your machine. Calls go straight from you to each provider. Nothing is proxied, pooled, or phoned home.

No payment method, ever

There is no card form anywhere. Not at download, not at sign-in, not hidden behind a feature. You sign in inside the app once and everything is included.

$0 by design

The signup agent runs on one free Gemini key you grab once, in about a minute, with no card. No paid CAPTCHA solvers, no SMS services, no subscription hiding anywhere.

It improves itself

FreeTokens uses the free models it harvested to discover new providers and verify new models. A human reviews every improvement before it ships to you.

06 — FAQ

Fair questions.

Is it really free?

Yes. There's no paywall, no trial clock, and we never ask for a credit card or any other payment method. Every key it collects is a real free tier from the provider itself. The Free Token Agent that does the signups runs on a free Gemini key you grab once, in about a minute, no card there either.

Do I need an account or a credit card?

You sign in once inside the desktop app, and that's it. No payment method is ever attached to your account: there is nothing to cancel, nothing to forget, nothing hidden behind an upgrade. Your API keys still live only on your machine.

Where do my API keys live?

On your machine, under ~/.freetokens. The gateway, database, and vault all run locally, and API calls go straight from your machine to each provider.

What happens when a signup needs a CAPTCHA or SMS code?

The agent pauses and asks you, in-app. You do just that one step (tick the box, paste the code) and it picks up exactly where it left off. 13 of the 24 providers need no help at all.

Is this allowed by the providers?

You're creating real accounts with your own name and email, on genuinely free tiers, with you in the loop. It is the same signup you'd do by hand, minus the tab-juggling. You're responsible for each provider's terms, and the app never pools or shares keys.

Do I need to be a developer?

No. Chat, document Q&A, agents, image generation, and voice are built into the app. If you are a developer, point Cursor, Claude Code, Aider, or any OpenAI-compatible client at http://localhost:8787/api/v1 with your lovie_ key.

Which platforms does it run on?

macOS (.dmg) and Linux (.AppImage) today; the Windows build is on the way.

Stop paying for tokens
you can get for free.

One download, one button, and every free tier worth having is in your vault. No credit card, no subscription, no payment method. Ever.