brynq 1.4.0

Python SDK for Intelligent Multi-Model Orchestration. Auto-routing, chain strategies, and local-first execution.

Brynq

All your AI models. One app.

---

Brynq connects the AI tools you already use — Gemini, Claude, ChatGPT, Llama, Mistral — and makes them work together. Log in with Google, and Brynq automatically finds every model on your machine.

No new subscriptions. No API keys to manage. Your AI, orchestrated.

You: "Research the latest on solid-state batteries and summarize the key breakthroughs"

  Gemini:  Searching recent papers and news... Found 12 key developments
           in 2025-2026 including Toyota's sulfide electrolyte breakthrough...

  Claude:  Analyzing Gemini's findings. The most significant trend is the
           shift from oxide to sulfide electrolytes, which reduces interface
           resistance by 10x...

  Llama:   Summary: Three breakthroughs matter most — (1) Toyota's sulfide
           cells hitting 500 Wh/kg, (2) QuantumScape's solid-state pouch
           cells passing automotive qualification, (3) Samsung SDI's...

---

Get started

Download the desktop app

Install for Windows   |   Install for macOS / Linux

Or install via terminal

pip install brynq

brynq-runtime start     # Opens the app
brynq-runtime chat      # Terminal chat
brynq-runtime models    # See all detected models
brynq-runtime status    # Check connections

---

How it works

1. Log in with Google

One click. Uses your existing Google account to access Gemini. No API keys, no credit card, no extra charges.

2. Brynq detects your models

Have Claude installed? Ollama running? ChatGPT CLI? Brynq finds them automatically and adds them to your team. Nothing to configure.

Model	How Brynq finds it	Cost to you
Gemini	Google OAuth (one-click login)	Your existing Google account
Claude	Auto-detected if installed on your machine	Your existing subscription
Llama, Mistral, Phi	Auto-detected via Ollama	Free (runs on your hardware)
ChatGPT	Auto-detected or paste API key	Your existing subscription or API key
Any Ollama model	Auto-detected	Free

3. They work together

You ask once. Brynq figures out which model handles which part, then executes locally on your machine. Five strategies:

• Sequential — One model's output feeds the next
• Fan-Out — Multiple models answer in parallel, best response wins
• Debate — Models critique each other's answers
• Refine — One drafts, another improves, repeat
• Single — Route to the best model for this specific task

---

Why Brynq

No extra charges — Uses the subscriptions and tools you already have. Google OAuth uses your Google account. Claude uses your existing installation. Ollama is free. No surprise bills.

Private by default — Local models run entirely on your computer. Your conversations never leave your machine. Cloud models are called directly from your machine to the provider.

Smarter together — Models fact-check each other and build on each other's strengths. Gemini is great at research, Claude at analysis, Llama at fast summaries. Together they produce better results than any single model.

One conversation — No more opening Claude in one tab, Gemini in another, and ChatGPT in a third. Ask once, get answers from your whole team, in one place.

---

Security

• Your keys stay local — API keys and OAuth tokens are encrypted on your machine. Never sent to our servers.
• Your data stays local — Documents and conversations never leave your computer.
• Encrypted storage — All credentials stored with Fernet encryption in a machine-bound vault.
• Tamper-proof — Execution plans are cryptographically signed and time-limited.

---

Development

# Install in dev mode
pip install -e ".[all]"

# Start the runtime
brynq-runtime start

# Run tests
python -m pytest tests/ -v

# Lint
ruff check src/brynq/ cloud/ runtime/

Stack

• Runtime: Python, FastAPI
• Desktop app: React 19, Vite 8, Tailwind v4, TypeScript
• Auth: Google OAuth 2.0 + BYOK API keys
• LLM support: Ollama, Claude, Gemini, OpenAI, LM Studio
• Tests: 4,500+ passing
• CI/CD: GitHub Actions

---

License

Apache 2.0

Built by the Brynq Team.