Works with every AI
Skales is the neutral ground: bring a key from any provider, sign in with the ChatGPT subscription you already pay for, or run fully offline with local models. Switch mid-conversation. Never get locked in.
Claude Fable, Opus, Sonnet & Haiku families
Bring your Anthropic API key and Skales drives Claude through chat, background goals, Code Mode and Codework. Power users can bring a Claude Code token. Claude's tool-calling discipline makes it a favorite for long autonomous runs.
www.anthropic.com ↗GPT-5 family, o-series reasoning, DALL·E, Whisper, TTS
Use an OpenAI API key - or sign in with the ChatGPT subscription you already pay for (Plus, Pro, Business, Enterprise), no API key needed. OpenAI also powers optional voice in/out (Whisper STT, 6 natural TTS voices) and Studio image generation.
ChatGPT subscription sign-in means no extra API costs.
openai.com ↗Gemini 3 Pro/Flash families
Google AI Studio keys drop straight into Skales. Gemini's long context is great for document-heavy goals, and the free tier is one of the most generous ways to start.
Real free tier via Google AI Studio.
ai.google.dev ↗Grok 4 family
Connect xAI's API and Grok answers inside Skales chat, goals and coding - with the same approval gates and local data storage as every other provider.
x.ai ↗DeepSeek V4 Pro & Flash (1M context)
First-class DeepSeek support, direct API or via the HuggingFace router. Agent-tuned V4 models are a community favorite for cost-effective autonomous work, and a built-in LLM Profile keeps their tool-calling sharp.
Among the cheapest capable agent models available.
www.deepseek.com ↗Kimi K2 family, 256K context, plus the Moonshot v1 models
Built in as its own provider: paste a key, pick International or China, done. Those are two separate accounts with separate keys, so the region is a switch rather than a URL you have to know - and the model list refreshes against the region you picked, which is how a newly released Kimi shows up without waiting for a Skales update. Kimi is one of the strongest open agentic models for long tool-using runs, and a built-in LLM Profile keeps its tool calls clean.
platform.moonshot.ai ↗Mistral Large, Codestral, open-weight models
European models with a real free tier. Skales ships a Mistral LLM Profile so even the smaller open-weight variants handle tools reliably.
Free tier available - and EU-hosted options.
mistral.ai ↗Llama, Qwen, Gemma, Mistral, GLM, Kimi - any local model
The 100% offline path: Skales auto-detects Ollama, browses the model library with one-click installs, and tunes small models with LLM Profiles and a tool-budget slider. No API key, no internet, no data leaving the machine. This is what local-first means.
Completely free. Your hardware is the only limit.
ollama.com ↗Any GGUF model with a friendly GUI
Run LM Studio's local server and Skales connects as an OpenAI-compatible endpoint - the easiest way to pair a model-management GUI with an agent that actually does work.
Free, local, no key.
lmstudio.ai ↗One key, 400+ models incl. free ones
One API key unlocks the whole model market - including genuinely free models to get started. Perfect with Skales' Fallback Provider Chain: define a priority list and outages never stop your goals.
Free models available on the standard tier.
openrouter.ai ↗Llama & friends at extreme speed
Groq's LPU inference makes Skales feel instant - and its free tier also covers Whisper speech-to-text, which Skales uses for voice input out of the box.
Generous free tier, including STT.
groq.com ↗Plus Cohere, Fireworks, Together, Nebius, Aleph Alpha, Perplexity, Cloudflare Workers AI, NVIDIA NIM, Cerebras, MiniMax and Hugging Face - and any OpenAI-compatible endpoint (KoboldCpp, vLLM, text-generation-webui). If it speaks the API, Skales speaks to it.
Start at $0
01
Install Ollama or LM Studio, pick a model in Skales’ built-in marketplace, done. No key, no internet, no data leaving the house.
02
Google AI (Gemini), Groq, OpenRouter free models, Cerebras and Mistral all have real free tiers - paste one key and go.
03
Sign in with ChatGPT Plus/Pro/Business in Settings → Subscriptions. The plan you already pay for becomes your agent’s brain.
Community-maintained list of current free tiers: Free LLM API Resources ↗
Model questions
Yes, two ways: run local models for free with Ollama or LM Studio (no key, no internet required), or use a free cloud tier - Google AI (Gemini), Groq, OpenRouter free models, Cerebras and Mistral all offer real free tiers you can paste into Settings → AI Providers.
Yes. Sign in with your ChatGPT account (Plus, Pro, Business or Enterprise) under Settings → AI Providers → Subscriptions - no API key needed. Power users can also bring a Claude Code or Gemini CLI token.
The Fallback Provider Chain switches to a backup provider automatically when your primary fails. You configure the priority order across all your providers - an outage or rate limit never blocks your work.
Yes - LLM Profiles (opt-in) tune the tool budget, prompt size and a per-model hint for DeepSeek, Qwen, Llama, Gemma, Mistral, GLM, Kimi and small local models, so weaker models stop fumbling tool calls. A guardrail warns when a ≤3B model is paired with too many tools.
