# opensandbox-group/OpenSandbox

Secure, Fast, and Extensible Sandbox runtime for AI agents.

Repository: https://github.com/opensandbox-group/OpenSandbox
Canonical: https://ross.abutalabs.com/products/opensandbox
Homepage: https://open-sandbox.ai
Language: Python
License: Apache-2.0
License Family: permissive
Topics: ai, ai-infra, kubernetes, sandbox, ai-agent
Last push: 2026-08-26T13:56:41+00:00

## Health v2 (maintenance only)
Score: 79/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 99, release rhythm 87, longevity 18
- inputs: {"age_days": 259, "days_push": 7, "days_rel": 7, "gap_med": 0, "n_releases_24m": 176}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 14737, forks 1321 (observed 2026-08-28T04:11:07.568821+00:00)

## What it is
OpenSandbox is a general-purpose sandbox platform for AI applications that securely runs commands, code interpreters, browsers, and developer tools in isolated environments. It provides multi-language SDKs (Python, JS/TS, Go, Java/Kotlin, C#), a CLI, MCP server integration, and Docker/Kubernetes runtimes with standardized lifecycle and execution APIs.

## Use cases
- run AI-generated code safely in an isolated sandbox
- execute coding agents like Claude Code or Codex in isolated environments
- run browser automation workloads with Playwright in a sandbox
- host a code interpreter for LLM tool use
- run agent evaluations and RL training environments at scale
- provide secure remote development environments in the cloud

## When to choose
- you need secure, isolated execution of untrusted or model-generated code
- you want to run coding agents or browser automation in reproducible sandboxes
- you need to scale sandbox workloads across a Kubernetes cluster
- you want multi-language SDKs and MCP integration for agent tooling

## When to avoid
- you only need simple local containerization without sandbox lifecycle management
- you need a fully managed hosted sandbox service rather than self-hosted infrastructure
- your workloads are unrelated to AI agents or code execution

## Facets
- artifact type: service
- maturity: active
- function: security, container-runtime, container-orchestration, sdk, mcp, cli, developer-tools
- domain: large-language-models, developer-tools, cloud-computing, self-hosted
- platform: python, go, jvm, dotnet, self-hosted, cross-platform
- tags: sandbox, code-execution, code-interpreter, ai-agents, isolated-environments, browser-automation, rl-training, coding-agents, containers, docker, kubernetes, nodejs

## Member repositories
- opensandbox-group/OpenSandbox (main) score 79

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:07.568821+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 2026-08-29T17:07:13.283401+00:00, confidence not recorded.
  - readme: https://github.com/opensandbox-group/OpenSandbox (fetched 2026-08-28T04:11:07.568821+00:00, sha fac6b42ea59c)
  - homepage: https://open-sandbox.ai (fetched 2026-08-29T08:05:20.771615+00:00, sha db9a0a05fec2)
  - site_page: https://open-sandbox.ai/getting-started (fetched 2026-08-29T08:05:20.780634+00:00, sha e764c36d9137)
  - registry_pypi: https://pypi.org/pypi/opensandbox/json (fetched 2026-08-29T08:05:20.782695+00:00, sha e95e1802d6d0)
- Data as of 2026-08-30T08:39:29.467469+00:00.