# akitaonrails/ai-jail

Multi-OS sandbox to run AI agents with better constraints (it is not 100% secure, but enough)

Repository: https://github.com/akitaonrails/ai-jail
Canonical: https://ross.abutalabs.com/products/ai-jail
Language: Rust
License: GPL-3.0
License Family: copyleft
Last push: 2026-09-02T18:18:55+00:00

## Health v2 (maintenance only)
Score: 83/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 100, release rhythm 100, longevity 13
- inputs: {"age_days": 185, "days_push": 0, "days_rel": 0, "gap_med": 1, "n_releases_24m": 54}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1172, forks 105 (observed 2026-09-03T02:15:08.910646+00:00)

## What it is
ai-jail is a Rust CLI that runs AI coding agents inside an OS-level sandbox using bubblewrap with Landlock, seccomp, and resource limits on Linux, and sandbox-exec on macOS. It provides secure defaults such as a private tmpfs home, opt-in agent credentials, and capabilities like network and GPU disabled by default.

## Use cases
- run claude code in a sandbox
- sandbox AI coding agents safely
- isolate AI agent filesystem access
- prevent AI agents from accessing my home directory
- run untrusted AI agent commands with restricted network
- sandbox agent credentials on macos and linux

## When to choose
- you run AI coding agents like Claude and want OS-level isolation without a VM
- you want secure defaults like private tmpfs home and opt-in credentials
- you work on Linux or macOS and want a lightweight sandbox layer

## When to avoid
- you need to run fully hostile code - use a disposable VM instead
- you are on Windows without WSL2
- you need guaranteed 100% security - the tool itself states it is not fully secure

## Facets
- artifact type: cli-tool
- maturity: active
- function: security, cli, developer-tools
- domain: security, developer-tools
- platform: cli, rust
- tags: sandboxing, bubblewrap, landlock, seccomp, ai-coding-agents, isolation, sandbox, ai-agents, command-line, linux, macos

## Member repositories
- akitaonrails/ai-jail (main) score 83

## Provenance
- Observed fields: from GitHub, fetched 2026-09-03T02:15:08.910646+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:36:45.955331+00:00, confidence not recorded.
  - readme: https://github.com/akitaonrails/ai-jail (fetched 2026-09-03T02:15:08.910646+00:00, sha 884856b54332)
  - registry_crates: https://crates.io/api/v1/crates/ai-jail (fetched 2026-08-29T12:41:38.507675+00:00, sha fba5c87342d1)
- Data as of 2026-08-30T08:39:29.467469+00:00.