OmniLLM CLI

A multi-provider LLM CLI agent (Claude Code style). BYOK. No backend. No accounts.

Automatically routes to the best available free LLM provider. Switches providers when rate limits hit. Compacts context when switching to smaller-window models.

Quick Start

# No install needed
npx omnillm

# Or install globally
npm i -g omnillm
omnillm

Add API Keys

/keys add groq gsk_xxxxxxxxxxxxx
/keys add gemini AIzaxxxxxxxxxxxx
/keys list
/keys test groq

Supported Providers (all free tier)

Provider	Best For	Models
Groq	Fast inference, coding	Llama 3.3 70B, 8B, Mixtral
Google AI Studio	Long context (1M), multimodal	Gemini 2.5 Flash
Cerebras	Batch processing	Llama 3.1 70B, 8B
SambaNova	Large models (405B)	Llama 405B, 70B, Qwen 72B
OpenRouter	Multi-model routing	DeepSeek R1, Llama 70B
GitHub Models	GPT-4o free	GPT-4o, Llama 3.3
NVIDIA NIM	Coding	DeepSeek R1, Llama 70B
Mistral	General purpose	Mistral Small
Fireworks AI	General purpose	Llama 3.3 70B
Ollama	Local (no key)	Any local model

Commands

Command	Description
/keys add <provider> <key>	Add an API key
/keys list	List all configured keys
/keys remove <provider> <id>	Remove a key
/keys test <provider>	Test key validity
/status	Quota and session dashboard
/model [provider] [model]	Force provider/model
/model auto	Auto provider selection
/compact	Manual context compaction
/help	Show help
/exit	Quit

How It Works

1. Classifies your task (coding, long context, fast reasoning, etc.)
2. Selects best provider among configured keys
3. Checks rate limits (RPM, TPM, RPD, TPD sliding windows)
4. Compacts context when switching to smaller models
5. Retries on failure with next-best provider

Storage

All data stored locally in ~/.config/omnillm/:

• keys.json — AES-256-GCM encrypted API keys
• quota.json — Daily usage counters
• config.json — User preferences

Environment Variables

See .env.example for all supported environment variables.

Development

npm install
npm run dev      # Run with tsx (no build needed)
npm run build    # Build single-file bundle