Any app.
One context.
Your IDE, your browser, your terminal, and your AI share a live context layer — without any of them knowing the others exist. No integrations. No plugins. No vendor lock.
The silo problem.
Today's AI tools each live in their own context window. Your IDE copilot doesn't know what's in your browser tabs. Your chat assistant doesn't know what's in your terminal. You spend your time copy-pasting between them, or re-explaining.
Strands removes that. Every connected app contributes to a single live context pool, routed through Conduit so only the relevant pieces reach the model on any given turn.
Strands reads from existing OS-level surfaces. No SDK, no plugin, no cooperation from the target app required.
Context updates as you work. When you switch tabs, the context layer knows. When you type in the terminal, the context layer has it.
Connected apps provide raw context. Conduit routes only what's relevant to the current query. The model never drowns in irrelevant material.
When Strands context reaches the model, every piece carries its origin. The proof system applies the same way as any other source.
Strands ships as part of Conduit. The open connectors (VS Code, Chrome, terminals) are in the public tier. Enterprise deployments add scoped connector policies, group-level context pools, and audit trails. Gov and Partner tiers add additional isolation and sealing guarantees.