Question 1

What is Xerant?

Accepted Answer

Xerant is a sandboxed multi-agent DevOps platform. It replaces a three-engineer on-call rotation with five specialized AI agents — Axiom (orchestrator), Forge (staff engineer), Warden (security), Vector (deployer), Sentry (observer). Each agent runs in its own gVisor sandbox and communicates only through a mediated bus, so a prompt injection in one agent cannot escalate to another. Free during beta.

Question 2

Is Xerant a Vercel or Netlify alternative?

Accepted Answer

Yes, but with a different scope. Vercel and Netlify host your app; Xerant hosts your deploy team. Xerant runs in your cluster, writes the PR, reviews it for security, ships the canary, and watches the rollout — work that Vercel and Netlify leave to your engineers. Teams using Xerant cut total hosting plus engineering cost by roughly 60% versus Vercel Pro plus a rotation.

Question 3

What is a gVisor sandbox and why does Xerant use one per agent?

Accepted Answer

gVisor is a Google-built user-space kernel that intercepts every syscall an application makes and forwards only the allowlisted ones to the host. Xerant gives each of its five agents a separate gVisor sandbox so that a compromised agent cannot read files, open sockets, or call tools that belong to a different agent. This is kernel-level isolation, not container-level — a container escape in Forge's sandbox still cannot reach Vector's deploy keys.

Question 4

How is this different from GitHub Agentic Workflows?

Accepted Answer

Xerant is a team, not an actor. Five agents with separate sandboxes and authority boundaries. Agentic Workflows runs one agent per workflow; we run five coordinated agents with mediated handoffs and independent rollback authority.

Question 5

What happens if an agent is prompt-injected?

Accepted Answer

It's contained. The gVisor sandbox blocks syscalls outside its allowlist, and the mediated bus blocks cross-agent calls that aren't in its role contract. Warden can be compromised and still not reach Vector's deploy keys.

Question 6

Can I self-host?

Accepted Answer

Yes, in the Enterprise tier. The full stack runs on your cluster; we ship the agent images and the bus.

Question 7

Which LLMs do you use?

Accepted Answer

Claude for Axiom, Forge, and Warden. A smaller model (configurable) for Vector and Sentry where determinism matters more than reasoning.

Question 8

What if I don't trust it to deploy?

Accepted Answer

Every agent can be set to advisory mode. Xerant drafts the PR, runs the review, stages the deploy — a human approves. Autonomy is a dial, not a switch.

Question 9

How do I try it?

Accepted Answer

Sign in with GitHub at xerant.vercel.app/signin. We'll give you a sandbox repo and you'll watch the full team ship a PR in under 10 minutes. Free while we're in beta — no card, no seat limits.

Welcome to Xerant.