Plimsoll, a self-hosted sandbox for code your AI agent writes - AI Coding Assistants

What is Plimsoll, a self-hosted sandbox for code your AI agent writes?

Plimsoll, a self-hosted sandbox for code your AI agent writes is A sandbox service for running untrusted, agent-authored code that reports the isolation boundary each run executed behind. Lessons, no clone needed: https://plimsollmark.github.io/plimsoll/trainers/ - plimsollmark/plimsoll It's designed for developers building agents or mcp tools that execute model-written code and want auditable isolation and credential-free api access.

About Plimsoll, a self-hosted sandbox for code your AI agent writes

Key Strengths

Every run reports the isolation tier it actually got (process, container, kernel or vm), and a request can set a minimum tier so nothing runs on a weaker sandbox
Capability grants let agent-written code call your API through a broker that attaches a per-run credential, so the sandbox never holds your key or open network access
One self-hosted Go daemon switches backends by config: WebAssembly/QuickJS or Docker locally, gVisor, E2B or Docker Cloud microVMs in production, with clients for Vercel AI SDK, Google ADK, CrewAI, Agno, Mastra and Trigger.dev

Limitations to Consider

Very young project: about 25 GitHub stars and one fork when checked, built by one developer for their own projects
The strongest (vm) tier depends on E2B or Docker Cloud, which bill per run; self-hosted kernel isolation needs gVisor installed
It sandboxes code an agent writes, not the agent itself; the README also says the reported tier is daemon evidence, not hardware attestation

Best For

Developers building agents or MCP tools that execute model-written code and want auditable isolation and credential-free API access.

Skip it if: You want a hosted, zero-ops code interpreter, or you need to sandbox a whole agent and its file system rather than individual code runs.

Pricing

Free

a completely free service with no subscription required

Free and open source (Apache-2.0); optional E2B or Docker Cloud backends bill separately per run. (checked Oct 13, 2026)

Sources

Pros, cons and pricing compiled from public sources (official pricing/docs pages, published reviews, user reports), not hands-on testing.

Compare Plimsoll

See 8 Plimsoll alternatives

Similar Tools in AI Coding Assistants

ST

Stroq

a local audit of what your coding agent did and why

Try Stroq

SP

Specification First Agentic Development Plugin and Prove

Spec-first agentic development is specification-driven development for AI coding agents: reviewed specs, isolated worktrees, gated commits.

Try Specification First Agentic Development Plugin and Prove

GR

Graphicle

Code graph and memory for your coding agents, in one Rust program. For Claude Code, Codex, Cursor and any MCP client. - Graphicle-AI/graphicle

Try Graphicle

Frequently Asked Questions

What is Plimsoll, a self-hosted sandbox for code your AI agent writes?

Plimsoll, a self-hosted sandbox for code your AI agent writes is A sandbox service for running untrusted, agent-authored code that reports the isolation boundary each run executed behind. Lessons, no clone needed: https://plimsollmark.github.io/plimsoll/trainers/ - plimsollmark/plimsoll It's designed for developers building agents or mcp tools that execute model-written code and want auditable isolation and credential-free api access.

How much does Plimsoll, a self-hosted sandbox for code your AI agent writes cost?

Plimsoll, a self-hosted sandbox for code your AI agent writes offers a completely free service with no subscription required You can get started for free on their official website.

Who should use Plimsoll, a self-hosted sandbox for code your AI agent writes?

Plimsoll, a self-hosted sandbox for code your AI agent writes is best suited to developers building agents or mcp tools that execute model-written code and want auditable isolation and credential-free api access.