Start here

You start an agent. You walk away from the laptop. The agent hits a permission prompt and pauses. Your phone buzzes. You tap Approve. The agent continues.

That's the entire product.

What it actually does

You install a small daemon on the machine where your agent runs. The daemon watches the agent's session and ships metadata (not transcripts) to a control plane. From your phone or browser anywhere in the world, you see what the agent is doing, and when it pauses for a decision, you can approve, deny, or steer. Without opening the laptop the agent is running on.

When the agent does risky work, it pauses for a plan review before touching the codebase. You read the plan on your phone. You approve. The agent moves on.

Who it's for

• Developers working with coding agents (Claude Code, Codex, Aider, Gemini) who want to step away from the keyboard without losing the loop.
• Engineering teams who want shared visibility into what agents are doing across the team, with approval routing and audit logs.
• Security-conscious orgs who want the wire path inside their own infrastructure. The relay self-host gets you most of the way there today.

Who it's not for (yet)

• Teams without coding agents in their workflow. Viewport is a control plane for agents. If you aren't running them, there's nothing to control.
• Air-gapped enterprises. The control plane is hosted-only today. On-prem is on the roadmap. talk to us if you need it.

Pick a path

The next three pages diverge by what you're trying to do.

Or if you'd rather understand the model before you install anything:

• What is Viewport · the product pitch and the runtime shape
• Mental model · sessions, plans, decisions, context, workflows