Skip to content

about

Why moatbox exists

Coding agents are good enough to work on their own, but not safe enough to work on your laptop with your keys. moatbox gives them a place to work where every tool call goes through one gate you control.

What it is

moatbox runs Claude Code, Codex and opencode headless, each in its own sandbox, inside a box that is yours. On the hosted plans every box is its own microVM on our servers; self-hosted, it runs on yours.

The tools are built in: Postgres, GitHub, Stripe, Slack, Gmail and more, with the real tokens kept outside the sandbox. Risky actions wait for your approval, supported ones can be undone, and every call lands on a timeline you can replay.

What we believe

  • Your keys, your models. You bring your own Anthropic or OpenAI keys or subscription. We never resell tokens.
  • Open source first. The whole product is Apache-2.0 and runs on your own server for free.
  • Approve before, undo after. An agent should ask before it drops a table, and you should be able to take it back.
  • Hosted in Europe. Hosted boxes run on our own servers in Germany and Finland.

Who builds it

moatbox is built by Valerio Massimiani, a developer turned founder based in Italy. Valerio builds software products, AI systems and web-extraction infrastructure, including webclaw, and writes about building them.

Say hello

Questions, ideas or a pilot for your team: write to [email protected].