aiagents.org

plimsoll

Self-hosted sandbox that runs agent-written code and reports its isolation tier.

  • AI tool

Visit site

Overview

plimsoll is a self-hosted sandbox that runs code that an AI agent produced, or other untrusted code. Every result states how strong the sandbox was, as one of four tiers: process, container, kernel, or vm. A request can require a minimum tier, and nothing runs if the sandbox falls short. It can also run an agent's program against hidden test scenarios and return SHA-256 hashes of each request and result, so a run can be checked later.

Key features

  • Four isolation tiers stated on every result
  • Requests can demand a minimum tier
  • Hidden test scenarios for checking agent-written programs
  • SHA-256 hashes of each request and result
  • Official Go client that rechecks the tier

Use cases

Alternatives

  • a0.devAI coding agent for building, deploying, and monetizing mobile appsai agent
  • AgemoAI software company building CodeWords for creating complex automationsai agent
  • Agent Chrome RelayLets coding agents use your signed-in Chrome in background tabs.AI toolFree
  • Agent StandupShows coding agent sessions as a live animated team call.AI toolFree
  • AgorShared boards where teams run coding agents on isolated git branches.AI tool
  • AideAI-native developer tool that autonomously works through software engineering tasksAI tool

Last updated 2026-10-15