Until now, seeing Lens Agents meant talking to us. That is the right call for a platform built for production and enterprise governance — but it's a slow way to answer the most basic question a builder has: what does this actually do?
So we wrote it down and made it runnable. You can now install the full Lens Agents platform on your own machine, launch a real agent, and chat with it in your browser — no sales call, no waiting list. Two guides, two paths, both live today:
- Try it with a coding agent — hand the setup to Claude Code, Codex, or OpenCode and let it drive.
- Try it by hand — the step-by-step Helm walkthrough, if you'd rather see every command.
What you actually get
This isn't a hosted demo or a screen recording. It's the real platform, installed via Helm into a dedicated local single-node cluster — its own minikube profile, isolated from any clusters you already run. When it's up you have:
- The Lens Agents platform — API, auth, policies, and audit — running on localhost.
- Prism, Lens Agents' ready-made managed agent, running in its own governed sandbox.
- A browser chat UI where you can put Prism to work.
In other words: the same platform organizations install into their own environments for production, running on your laptop in an afternoon.
Pick your path
Let a coding agent do it. Open Claude Code, Codex, or OpenCode and paste one prompt. It installs the Lens Agents admin skill, stands up the platform, launches Prism, and hands you back both URLs — pausing only when it needs you to sign in. This is the fastest self-serve path we've ever shipped.
Or do it by hand. The manual guide walks every command: start the cluster, helm install the platform from GHCR, expose it, activate with your Lens ID, create a policy, and launch Prism. Same result — you just see the whole machine.
Either way you'll need a few standard tools on the host (docker, minikube, kubectl, helm 3.8+, openssl, curl), a free Lens ID, and one inference-provider credential — an AWS Bedrock API key, or a Microsoft Foundry URL and key with a Claude deployment. The setup prompts you for anything that's missing.
Straight talk on the local trial
A laptop can't reproduce a production security boundary, so we won't pretend it does. This local setup shares the host kernel: agents run with container-level isolation — privilege dropping, kernel-level network lockdown (nftables), and proxy-mediated egress — the standard sandbox harness, but not microVM isolation.
Production deployments add per-sandbox microVM isolation with a Kata Containers or gVisor runtime class, which a local single-node cluster simply can't provide. The local trial is the honest way to understand Lens Agents — how policy, credential mediation, audit, and managed inference fit together. When you're ready to scope a production-grade environment, talk to us.
Why this matters
Every team wants AI agents. Today most of them run with personal credentials, inconsistent sandboxing, and data flowing to model providers uncontrolled. Lens Agents is the governed platform for running those agents at organizational scale — any agent, any model, your infrastructure — with identity, policy, credential mediation, and a full audit trail built in.
The fastest way to understand that is to run it. Now you can.