Run polished local models
Embedded Pi agent harness for knowledge work, built around OpenAI’s gpt-oss-20b, with plugin support for reusable, skill-based workflows.
A private, collaborative AI assistant that works for you. Built with local models and AT Protocol.

For sensitive knowledge work or personal tasks, your AI shouldn’t require you to trust a third party with your conversations and intellectual property. Run models locally, sync chats peer-to-peer with end-to-end encryption, and use social features, such as sharing chats, built on AT Protocol. Tiles gives you user-owned keys for your digital life with DIDs and UCANs, so your data and identity truly remain yours.
Embedded Pi agent harness for knowledge work, built around OpenAI’s gpt-oss-20b, with plugin support for reusable, skill-based workflows.
Tiles uses locally generated DIDs and UCANs for zero-trust authentication and authorization. This lets you own and control your identity across your digital life.
Securely run local models on a remote device like your home lab, peer-to-peer, from any device.
Encrypted peer-to-peer chat sync across your linked devices.
Create a public or private link to a chat session, published on Atmosphere and stored on your own personal data server (PDS).
Tiles is open source and forever free to use, dual-licensed under MIT and Apache 2.0. It is designed to be future-proof and continue working forever.
Plan finances, track health notes, and keep a private journal. Run models locally so that context stays on your machines, sync across your devices, and share only when you choose.
Turn shared compute workstations into private AI infrastructure. Members connect from their own laptops, use the space's local models, and collaborate while keeping their identity and data theirs.
Review literature, analyze datasets, and launch AI agents in preconfigured, resumable sandboxes with the right research tools already installed.