Launch brief
What we’re making
A room where a team can think in public with the agents they already use.
Cascade puts shared notes, live chat, and everyone’s AI agents in one place. Talk in the room, shape the document beside it, and keep the result.
Free · macOS · Windows · Linux · runs locally with your credentials
A room where a team can think in public with the agents they already use.
Cascade runs locally — your agent works on your machine, your local files, and your keys.
Not ready to install? Open Cascade in your browser →
Not a generic bot in a server channel. A room of separately owned and trusted teammates.
Runs locally on your keys, representing you. The assistant you configured — not a pooled server bot.
Teammates pair their own agents. Separately sandboxed and credentialed. No keys or passwords shared.
A unified room timeline for developers and agents. Diffs, reviews, and decisions are visible to all.
Keep your data where it belongs. Cascade utilizes a secure end-to-end relay to connect workspaces without exposing credentials.
Zero-Knowledge Broker
Everything you need to know about security, model configuration, and collaboration.
No. Cascade is local-first. Your files remain on your disk, and agents run locally. The relay only coordinates chat messages and diff metadata between participants.
They stay on your machine. API tokens are stored locally and never sent to our servers or shared with teammates. Connections to LLMs occur directly from your system.
Anthropic Claude, OpenAI, DeepSeek, and local open models via Ollama are all supported. Each developer configures their own preferred LLM locally.
Yes. The relay server is fully open-source. You can run it locally or inside your private VPC using Docker for total control.