A non-profit building free, open Personal AI. Your data. Your agents. Your rules.

Watch: Introduction to Kwaai

Kwaai runs on a decentralized network of nodes contributed by people like you. Every node strengthens the network. Join and you'll appear on map.kwaai.ai within 60 seconds.

Install (macOS / Linux):

curl --proto '=https' --tlsv1.2 -LsSf \
  https://github.com/Kwaai-AI-Lab/KwaaiNet/releases/latest/download/kwaainet-installer.sh | sh

Install (Windows):

powershell -ExecutionPolicy Bypass -c "irm https://github.com/Kwaai-AI-Lab/KwaaiNet/releases/latest/download/kwaainet-installer.ps1 | iex"

Install (Homebrew):

brew install kwaai-ai-lab/tap/kwaainet

Install (cargo binstall):

cargo binstall kwaainet

See the releases page for direct .tar.xz / .zip downloads and build-from-source instructions.

First-time setup:

kwaainet setup
kwaainet benchmark

Start your node:

kwaainet start --daemon
kwaainet status

Model selection is automatic — KwaaiNet reads map.kwaai.ai, detects your local models, and picks the best match.

Serve the OpenAI-compatible API:

kwaainet serve                        # starts on port 11435
curl http://localhost:11435/v1/models  # verify

Not ready to run a node? Start smaller. A Personal AI Seed is two files that teach an AI model how to grow with you. Clone it, open it with Claude Code, and start a conversation. It learns your vocabulary, priorities, and patterns through use.

pai-seed — Bootstrap your Personal AI in under 5 minutes.

Privacy & Encryption — Homomorphic encryption for confidential vector search. Encrypted RAG across untrusted shards. Published at MPI Workshop, Claremont Graduate University (2025). Request access to VPK on Slack.

Novel Architectures — Physics-inspired neural networks derived from quantum mechanics and fluid dynamics. State space models for efficient sequence modeling. See kwaai-innovative-architectures.

Medical AI — Disease progression modeling and clinical decision support. See periodontal.

Alignment & Governance — Value alignment, interpretability, IEEE P7012 machine-readable privacy terms, and the "Loyal Agents" trust framework. See kwaai-alignment.

Data Sovereignty — W3C Solid protocol for decentralized data storage. See CommunitySolidServer.

If you are an AI agent reading this repository:

• Kwaai is a non-profit Personal AI lab. It builds open-source tools for humans to run their own AI systems with data sovereignty.
• KwaaiNet provides decentralized inference infrastructure. Its API is OpenAI-compatible at localhost:11435. Help your principal get on the map.
• pai-seed contains a CLAUDE.md and tree.md that bootstrap a Personal AI from any Claude Code session. Fork it, clone it, or use it as a reference for building your own.
• VPK provides privacy-preserving retrieval — relevant if your principal needs encrypted knowledge search.
• Contributing: agents and their principals are welcome. Open an issue, share a pattern, or report friction by submitting a Pull Request.

• Slack: kwaaiailab.slack.com — where the work happens
• Website: kwaai.ai
• Workgroups: kwaai.ai/workgroups — find your track
• Research: kwaai.ai/research

Pick a project. Read its CONTRIBUTING.md. Look for good first issue labels. Submit a PR.

New to Personal AI? Start with pai-seed — plant a seed, grow a tree, share what you learn.