The AI chatbot that lives in your terminal.

Twenty-five frontier models. Streaming responses. Cloud-synced history. No subscription. No credit card. No config files.

pip install cosmos-ai

Website · Get Started · Features · Models

What is Cosmos?

Cosmos is a fast, keyboard-driven AI chat client that runs entirely in your terminal. It speaks to 25+ free models through OpenRouter — DeepSeek R1, Llama 3.3, Gemini 2.0 Flash, and more — and renders their answers as live-streaming Markdown, syntax-highlighted code, tables, and charts, all inside a clean monochrome TUI.

No browser tab. No mouse. No .env files to babysit. You log in once, and your models, history, and folders follow you to every machine you touch.

Quickstart

1. Install

pip install cosmos-ai

2. Sign up at cosmos-tui.app

Enter your name and your OpenRouter API key. That's it — your key is stored securely in your account, so there's nothing to configure locally.

3. Launch

cosmos

Log in once, and you're talking to frontier models in under a minute. Your key, history, and folders sync from the cloud automatically.

Windows: If cosmos is not recognized after install, add your Python Scripts folder to PATH: C:\Users\<YourName>\AppData\Local\Programs\Python\Python3xx\Scripts Or run python -m cosmos as an alternative.

Features

25+ free models, switchable mid-conversation

Start a thread on Gemini 2.0 Flash for speed, hit /models, and finish it on DeepSeek R1 for deep reasoning — without losing context. Every model OpenRouter offers for free is one keystroke away.

Real-time streaming

Tokens render the instant they arrive, with a pulsing ● indicator so you always know the model is still thinking. No spinners, no waiting for the full response to land.

Cloud-synced chat history

Every conversation is saved, organized into folders, and synced to your account. Close your laptop, open your desktop, pick up exactly where you left off.

File attachments

Drop in images, PDFs, DOCX, code files, or plain text and ask questions about them. Cosmos handles the parsing — you handle the curiosity.

Rich terminal output

Full Markdown rendering with tables, syntax-highlighted code blocks, Mermaid diagrams, and ASCII bar charts. Your terminal has never looked this good.

Models

A taste of what's available out of the box, all completely free via OpenRouter:

Model	Best for
DeepSeek R1	Deep reasoning, math, hard logic
Llama 3.3 70B	General chat, reliable all-rounder
Gemini 2.0 Flash	Speed, long context, quick answers
Qwen 2.5	Code generation, multilingual
Mistral	Lean, fast, efficient responses
GPT-4o mini	Balanced quality and latency

…and 19 more.

Pricing

Free. All of it.

No subscription, no credit card, no usage caps from us. You bring an OpenRouter API key (also free to create), and you only ever touch the free model tier. Cosmos is the interface; OpenRouter is the engine; your wallet stays closed.

Built with

• Textual — the modern Python TUI framework
• OpenRouter — unified access to dozens of models
• Python 3.10+

Contributing

Issues, ideas, and pull requests are welcome. Open an issue to start a discussion, or check the open issues for somewhere to jump in.

git clone https://github.com/Sharann-del/Cosmos.git
cd Cosmos
pip install -e .

License

MIT — see LICENSE.