Your keys. Every AI model. One terminal.

Made for people who use more than one model

You already pay OpenAI, Anthropic or Google directly. kunji puts all of them behind one clean prompt, and keeps you in control of the bill.

Switch models without losing the thread

Type /model groq:llama-3.3-70b-versatile and keep talking. The conversation carries over.

See the cost of every reply

Tokens and price under each answer, /cost for the session, kunji usage for the week. In your currency, including ₹.

Daily budgets

Set a limit per provider or overall. kunji warns at 80% and stops at 100%, so a long night never turns into a surprise bill.

Compare models side by side

kunji compare "…" sends one prompt to several models at once and shows which was fastest and cheapest.

Saved prompts

kunji @commit writes a commit message from your staged changes. @explain, @review and @translate are built in, and you can add your own.

Chats you can come back to

Every conversation is saved on your computer. Continue the last one with kunji chat -c, or search them all.

See all features

Ten providers, plus any compatible server

Run local models with Ollama for free, or add LM Studio, vLLM or your own proxy with one command.

OpenAI, Anthropic, Google Gemini, Groq, OpenRouter, DeepSeek, Mistral, xAI, Together AI, Ollama and your own OpenAI-compatible server.

Set up a provider

Your keys never leave your machine

Keys are stored in Windows Credential Manager, macOS Keychain or Linux Secret Service, never in a text file. Your prompts go straight to the provider you chose. kunji has no server in between.

Start in a minute

Needs Python 3.11 or newer. Full guide: installation docs.

1. Install

   Works on Windows, macOS and Linux. If pip isn't found, run python -m pip install kunji.

   pip install kunji

2. Open kunji

   The menu helps you add a key. kunji checks it works, then keeps it in your OS keychain.

   kunji

3. Chat

   Pick Start a new chat. Type / to see every command.

   › hello