Code & AI agent sandboxes: read the guide

OpenSandbox

Open source toolSandbox

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

Homepage Docs Repository

Scanned 12 Oct 2026 · Schema v5 · features v21

Licence: Apache-2.0
Self-hosted; you pay for the infrastructure.
Free tier: Free to run
Isolation: Container
shares the host kernel
Regions: —
Not disclosed on the public surface.
Capabilities: 23 / 61
6 partial · 29 fail · 0 not assessed
Agent readiness: 5/10 · 9/10
read · drive, pre-login surface

Advertised for

Untrusted agent code · Shell command execution · Code interpreter · Browser automation · Desktop computer use · Coding agent environments · Agent evaluation · Rl training

Built for

Developers, then technical

· SDK-first lifecycle/execution APIs, source installation and Kubernetes/runtime configuration.
· Runnable coding-agent, browser and evaluation examples.

01 Runtime Each fact carries its source; dots show confidence. Open a row for the evidence.

Isolation & lifecycle

Isolation primitive

Containershares the host kernel

Persistence

Backend-dependent: host bind mounts, Docker named volumes, Kubernetes PVCs and OSSFS mounts; snapshots persist separately.

Snapshots / forks

Docker image snapshots; Kubernetes container rootfs OCI snapshots; FastSandbox reusable checkpoints with microVM device/vCPU/memory state.

Resources & limits

Storage

Host paths, Docker named volumes/PVCs, Kubernetes PVCs and Alibaba OSSFS mounts; snapshots/golden images use operator registries/object storage.

Limits & quotas

Creation timeout defaults to 10 minutes in high-level SDKs; configurable/null manual cleanup for image/snapshot creation. Template creation requires TTL. Optional creation-time timeout ceiling does not cap renewals or total lifetime. CPU/memory/GPU limits, cluster admission quotas and pool acquisition/create-wait deadlines depend on operator/backend.

GPU

Yes

Access

Deploy interface

http-apisdkclimcpdockerkuberneteshelm

SDK languages

PythonJavaScriptTypeScriptKotlinJavaGoC#

Network & secrets

Egress controls

Per-sandbox DNS rules plus opt-in dns+nft default-deny network enforcement, operator overlays and runtime policy changes.

Secrets

Environment injection plus optional sidecar Credential Vault on-the-wire injection; template-backed FastSandbox has no sandbox-side Vault.

Bring your own cloud

Placement & use

Jurisdiction

Advertised workloads

Untrusted agent codeShell command executionCode interpreterBrowser automationDesktop computer useCoding agent environmentsAgent evaluationRl training

02 Capabilities 23 pass · 6 partial · 29 fail · 61 of 61 assessed

Supported — Documented on the public surface and verified.
Partially supported — Exists with caveats: gated, thin, or enterprise-only.
Not supported — Checked and not offered.
Not assessed — Scanned before this capability was tracked. Not a lack.

Compute & storage 6/10

Persistent diskStorage that survives restart/stop (not ephemeral-only).

Supported

Object storageManaged S3-compatible / blob object store the platform PROVIDES to a user's workloads (buckets, presigned URLs). DISTINCT from persistent-disk (block/volume attached to an instance) and from a self-host dependency (source.self_host.deps[]=["object-storage"] is an infra dep to RUN an OSS tool — the opposite direction). SCOPE varies and matters: native first-class (Railway Buckets, Cloudflare R2) vs a sibling SKU / extension (Fly Tigris, Bunny Storage, DigitalOcean Spaces) — record which in the evidence quote / note. Admitted v15.

Not supported

Snapshots / forkingSnapshot or copy-on-write fork of a running instance (filesystem-only counts). Whether the snapshot also restores memory + running processes is recorded separately by memory-snapshot (v19).

Supported

Live-state (memory) snapshotsA snapshot / checkpoint / pause captures a running instance's MEMORY and PROCESS state and restores it, so running processes continue where they left off instead of a freshly booted instance with the same files. Counts whether user-triggered (snapshot, checkpoint, pause-to-storage) or used to restore pre-initialised state (memory snapshots for fast cold starts). Distinct from snapshots-forks (any snapshot or fork, including filesystem-only; a product passing that key may still fail this one) and from scale-to-zero (an idle-cost property; suspension may or may not keep memory). Actor multiplexing (see scale-to-zero) records its live-state half here — only when the suspended actor's MEMORY/process state is what comes back, not a freshly booted worker with the actor's files.

Supported

Scale to zeroIdle instances suspend at no / negligible cost. ACTOR MULTIPLEXING (many suspended stateful actors sharing a small pool of pre-warmed workers, woken on request — Agent Substrate, AX, kagent; Rivet's hibernating actors) has NO key of its own (proposal REJECTED v20) — score the idle-cost half HERE and the live-state half under memory-snapshot; the oversubscription ratio / density is a performance characteristic the registry does not key.

Partially supported

GPU computeGPU-backed instances available.

Supported

Multi-regionCompute placeable in more than one geographic region.

Not applicable

Sandbox templates / custom imagesBuild reusable sandbox/VM images from an OCI image or Dockerfile (a base you fork instances from).

Supported

Cloud-init / ignition configCustomise a VM / microVM / instance at boot via cloud-init (or ignition) user-data — the same boot-metadata interface as AWS/Azure/GCP. Inject packages, users, SSH keys, files, and run-commands declaratively at launch. Distinct from sandbox-templates (a PREBUILT image you fork instances from) — this is PER-LAUNCH declarative customisation, not a baked image.

Not supported

Desktop / computer-use sandboxA virtual desktop (GUI) for LLM computer-use — browser + desktop automation, often via noVNC/VNC.

Supported

Selectable isolation runtimeThe user chooses the isolation runtime per workload / environment / sandbox from more than one option (e.g. microVM vs gVisor vs plain container, native namespaces vs gVisor, gVisor vs VM). Records the CHOICE; the isolation_primitive facet still records the best-evidenced primary. A platform that switches runtime automatically without a user choice (e.g. gVisor forced for GPU) does not pass on that alone.

Partially supported

Networking & delivery 3/9

Automatic HTTPSTLS endpoints provisioned automatically per workload.

Not supported

Built-in reverse proxyPlatform routes external traffic to workload services.

Supported

Custom domainsBring-your-own domain via CNAME/DNS.

Partially supported

Public preview URLsShareable per-deploy / per-branch URLs.

Supported

Direct public IPv6Per-instance stable public IPv6 for raw TCP/UDP (beyond a managed HTTPS proxy) — e.g. mosh / raw sockets.

Not supported

Egress controlsConfigurable OUTBOUND network policy — egress firewall / allowlist / egress proxy on traffic LEAVING a workload. Scoped to OUTBOUND only (v15): private-network REACHABILITY between a user's own services is private-networking, not this. Static-egress-IP belongs here (an outbound concept).

Supported

Private networkingA platform-internal private network between a user's OWN services — service-to-service connectivity, VPC, service mesh, and/or service discovery, not exposed to the public internet. DISTINCT from egress-controls (OUTBOUND firewall/allowlist), reverse-proxy (PUBLIC inbound ingress), and direct-ipv6 (per-instance PUBLIC addressing). Admitted v15.

Partially supported

Email send / receiveSend and/or receive email from the workload.

Not supported

Edge KV storeA globally-replicated key-value store read from WITHIN an edge runtime at the PoP (low-latency, eventually-consistent) — Cloudflare Workers KV, Deno KV, Fastly KV Store, Akamai EdgeKV, Vercel Edge Config. Distinct from first-party-db (a managed DATABASE service you provision and connect to) and from http-query-api (HTTP access to an external store) — edge-kv is colocated with the edge compute and read at the edge.

Not supported

Developer access 5/10

Public APIDocumented programmatic API.

Supported

OpenAPI specMachine-readable OpenAPI / Swagger spec published.

Supported

CLIOfficial command-line tool.

Supported

SDKOfficial client library/libraries.

Supported

SSH accessDirect SSH into workloads.

Not supported

Filesystem APIRemote file operations over the API (list/read/write/watch/copy/upload) without SSH.

Supported

Git deployDeploy from a Git repo / push-to-deploy.

Not supported

Private container registryHosted private image registry.

Not supported

E2B-compatible APIDrop-in compatibility with the E2B sandbox API/SDKs — existing E2B client code runs against this product by changing the endpoint/domain + key, with no business-code changes. E2B itself passes as the reference implementation. Distinct from public-api / sdk (the product has its own API — this key says it also speaks another vendor's) and from self-host-oss (a self-hosted E2B-compatible server is scored on both).

Not supported

HTTP data APIQuery / read / write the stored DATA over plain HTTP/REST (or GraphQL) WITHOUT the native protocol driver (Postgres wire / Redis RESP) — agent- and edge-friendly data access via fetch(). The DATA plane; distinct from public-api (a control-plane / management API). Auto-generated-from-schema REST/GraphQL (PostgREST, pg_graphql) is a flavour of this.

Not supported

AI & agents 4/12

Built-in coding agentShips a first-party AI coding agent in the product.

Not supported

Agent skillsInstallable / declarable agent skills or tools.

Supported

AI app builderPrompt-to-app / vibe-coding generation surface.

Not supported

MCP serverExposes a Model Context Protocol server for agents.

Supported

MCP clientActs as an MCP client / consumes MCP tools.

Not supported

MCP-server hostingRun/host MCP servers as a workload on the platform (distinct from exposing a first-party MCP control server — that is mcp-server).

Not supported

LLM gatewayRoutes LLM API calls through the platform (multi-provider).

Not supported

BYOK for LLMsBring-your-own LLM provider keys.

Supported

Model inference API (first-party)A FIRST-PARTY hosted LLM/model inference API — the provider serves its OWN catalogue of (usually open-weight) models behind a (typically OpenAI-compatible) endpoint, billed per token. Distinct from llm-gateway (a multi-PROVIDER proxy/router for YOUR calls to external providers like OpenAI/Anthropic) and from byok-llm (bring your own provider keys) — here the platform IS the inference provider, serving its own models.

Not supported

Autonomous QA agentA first-party AUTONOMOUS AI agent that QA-TESTS a running app / implementation — exercising it like a real user (clicking through flows, checking behaviour) and reporting bugs/defects (often back to a coding agent, or before a human review). Distinct from built-in-agent (a code-AUTHORING agent — this one VERIFIES the result, it does not write the feature; a product can ship both), from ai-guided-code-review (organises a diff for a HUMAN to review — not an autonomous tester), and from observability (runtime metrics/logs, not active testing).

Not supported

Credential-injection proxyThe platform holds real secrets OUTSIDE the workload and injects them on the wire (an egress / broker proxy swaps a placeholder or phantom token for the real credential on matching outbound requests), so code running in the sandbox — e.g. an untrusted agent — never sees or can exfiltrate the real key. Distinct from secrets-management (storage + injection INTO the workload as env vars/files — the workload holds the secret) and from egress-controls (outbound allow/deny POLICY; a product can have an allowlist without injecting credentials, and vice versa).

Supported

Preinstalled coding agentThe platform installs and launches a THIRD-PARTY coding agent (a harnesses.yml entry — Claude Code, Codex, OpenCode, Pi ...) inside a workspace/sandbox it provisions, with the agent's auth wired in (your own account or API key), so you start an agent session rather than an empty box. Distinct from built-in-agent (a FIRST-PARTY agent the vendor ships) and from harnesses[] (which records that a harness CAN target the product, not that the product runs it for you).

Partially supported

Data 0/4

First-party databasePlatform-provided managed database.

Not supported

Database branchingCopy-on-write branch of a DATABASE (schema + data) as a dev-lifecycle primitive — branch your data like code, usually instant via a metadata operation. Distinct from snapshots-forks (forks a running compute INSTANCE / VM state) and from preview-environments (a full per-PR environment that may USE a DB branch). For a data product a CoW branch maps HERE, not to snapshots-forks.

Not supported

Vector searchFirst-class storage, indexing, and similarity search of vector embeddings — a native vector type/index or a managed vector DB. The data-layer primitive for RAG / semantic retrieval.

Not supported

Ephemeral instant DB (no-signup)Provision a working, usable instance via a single command / HTTP call with NO account, signup, or card — typically time-boxed (expires unless claimed). A zero-onboarding path an agent can hit directly.

Not supported

Platform & operations 2/6

Scheduled tasksCron / scheduled job primitive.

Supported

Durable workflowsDurable execution / long-running workflow primitive.

Not supported

Preview environmentsA full EPHEMERAL environment spun up per PR / branch — compute + addons + (often) a DB branch — torn down on merge. More than preview-urls (which is just a shareable URL for one deploy).

Not supported

Built-in observabilityFirst-party metrics / logs / tracing / APM / profiling included in the platform (no third-party stack required to see what a workload is doing).

Supported

Managed CI/CD runnersPlatform-hosted, drop-in CI/CD runner VMs (GitHub Actions / GitLab), typically faster and cheaper than the CI vendor's own hosted runners, billed per runner-minute; isolated + ephemeral per job. Distinct from git-deploy (push-to-deploy of an APP) and scheduled-tasks (cron) — this is managed CI COMPUTE you point your existing pipelines at.

Not supported

Project template marketplaceA marketplace / gallery of prebuilt PROJECT STARTER templates (apps or websites) a user clones or launches a new project from and then customises — the "start from a template" surface of a builder / PaaS. Distinct from sandbox-templates (a COMPUTE base IMAGE you fork runtime INSTANCES from — Dockerfile / OCI), from app-builder (the prompt-to-app GENERATION surface — a product can have both), and from one-click OSS-app catalogues / helm-addon-marketplace (deploy a FINISHED app as-is, not a starter you build on).

Not supported

Organisation & security 3/7

Built-in authPlatform-provided authentication / IAM for workloads.

Partially supported

SSO / SAMLEnterprise single sign-on.

Not supported

Secrets managementFirst-class secrets storage / injection.

Supported

TeamsMulti-user team accounts.

Not supported

SharingShare a running workload/resource via link (optionally gated).

Supported

BYOCRuns inside the customer's own cloud account.

Not applicable

Deploy into customers' VPCsA SaaS vendor uses the platform to deploy and operate ITS software inside ITS CUSTOMERS' cloud accounts / VPCs (nested or "customer BYOC"), managed from the vendor's control plane. Distinct from byoc (your own workloads in your own cloud account).

Not applicable

Workload identityThe platform gives each workload its own verifiable identity (OIDC token, SPIFFE SVID, cloud IAM federation) that it exchanges for short-lived credentials, so no long-lived static secret is stored or injected. Distinct from secrets-management (stores / injects static secrets) and credential-injection-proxy (swaps real secrets in on the wire).

Not supported

Self-host / OSS coreOpen-source core that can be self-hosted.

Supported

Proposed, not verified

Capabilities the scan saw but the registry does not track yet. A key is admitted only once two distinct products exhibit it. These are not scored.

Experimental client-side sandbox poolsandbox-pooling
“The OpenSandbox SDKs ship an experimental client-side sandbox pool that keeps a small buffer of ready sandboxes warm on the server so that acquire() returns quickly instead of paying the full sandbox creation latency on the hot path.” github.com
Per-execution Bubblewrap isolation sessions within one sandboxintra-sandbox-isolation-sessions
“Isolation sessions run a long-lived bash inside a bubblewrap namespace, so one sandbox pod can host many mutually isolated task runs without spinning up a new container. Each session gets its own PID, mount, tmpfs, and env namespaces; startup is sub-millisecond.” github.com

03 Agent readiness Pre-login surface · counts, not a grade

Can an agent read it?

parse-ability · docs, metadata, machine-readable surfaces

5/10

4 full · 1 partial · 50% of applicable checks

llms.txt
Not supported
“https://open-sandbox.ai/llms.txt: HTTP 404; no llms.txt source found in the repository documentation paths.”
llms-full.txt
Not supported
“https://open-sandbox.ai/llms-full.txt: HTTP 404.”
robots.txt AI policy
Partially supported
“https://open-sandbox.ai/robots.txt: HTTP 404, therefore default-allow, not explicit AI allow.”
JSON-LD
Not supported
“Homepage HTML contains no application/ld+json script.”
Sitemap
Not supported
“https://open-sandbox.ai/sitemap.xml: HTTP 404; absent robots.txt supplies no alternate sitemap pointer.”
Meta tags
Supported
“Homepage title, description and og:title/og:description describe OpenSandbox Documentation and Universal Sandbox Infrastructure for AI Applications.”
Markdown negotiation
Not supported
“Accept: text/markdown on the homepage still returns text/html; /getting-started/index.md returns 404. Public raw GitHub Markdown is available but is not website content negotiation.”
Content efficiency
Supported
“Homepage has 1,996 UTF-8 bytes of visible text / 31,015 HTML bytes = 6.44%, above the rubric’s 5% threshold.”
Docs discoverable
Supported
“Server-rendered homepage navigation links Getting Started, Architecture, Guides, SDKs, API and Deployment.”
Machine-readable pricing
Not applicable
“Self-host-only Apache-2.0 project has no first-party hosted pricing tiers.”
No anti-bot wall
Supported
“Homepage, raw source/docs and SVG assets were fetched without a public-content challenge.”

Can an agent drive it?

act-ability · APIs, SDKs, MCP, credentials, webhooks

9/10

8 full · 1 partial · 90% of applicable checks

Public API
Supported
“https://github.com/opensandbox-group/OpenSandbox/blob/main/docs/api/index.md links substantive lifecycle, execd, egress and diagnostics API documentation.”
OpenAPI spec
Supported
“Fetched lifecycle (OpenAPI 3.1.0), execd and egress source specifications from https://github.com/opensandbox-group/OpenSandbox/tree/main/specs.”
MCP server
Supported
“https://github.com/opensandbox-group/OpenSandbox/blob/main/docs/sdks/mcp.md documents pip install opensandbox-mcp and first-party lifecycle/command/file tools; /.well-known/mcp being 404 does not remove credit.”
SDK packages
Supported
“https://github.com/opensandbox-group/OpenSandbox/blob/main/docs/sdks/index.md documents five official SDK families with package installation commands.”
Auth documented
Supported
“https://github.com/opensandbox-group/OpenSandbox/blob/main/docs/guides/multi-tenancy.md documents tenant API-key authentication and scoped access; secure-access documents endpoint tokens.”
Self-serve credentials
Not applicable
“There is no first-party hosted account/key-issuance service. Self-host operators configure API keys; no live credential acquisition was attempted.”
Webhooks
Not supported
“No documented signed outbound webhook integration found in the captured architecture/API/SDK surface; in-sandbox hooks are not webhooks.”
CLI tool
Supported
“https://github.com/opensandbox-group/OpenSandbox/blob/main/cli/README.md documents official osb/opensandbox-cli, machine-readable output and diagnostic commands.”
Error model
Supported
“https://github.com/opensandbox-group/OpenSandbox/blob/main/specs/sandbox-lifecycle.yml defines ErrorResponse required code/message fields, HTTP categories and request IDs.”
Rate limits documented
Partially supported
“Lifecycle API documents 429 pool-capacity timeout plus Retry-After; server configuration documents Kubernetes read/write QPS knobs. No caller-facing rate-limit/quota schedule is documented.”
Agent config files
Supported
“Repository AGENTS.md, troubleshoot-sandbox/SKILL.md and osb skill listings/install targets are multiple documented agent-configuration surfaces.”

04 Coding harnesses Which agents and IDEs are documented to run here

Advertised integrations 11

Claude Code Anthropic
Supported
“Access Claude via the `claude-cli` npm package in OpenSandbox.”
Source github.com
First-party documented runnable integration example; not behaviorally exercised during this scan.
Codex CLI OpenAI
Supported
“Use the official @openai/codex npm package to call OpenAI/Codex-like models in OpenSandbox.”
Source github.com
First-party documented runnable integration example; not behaviorally exercised during this scan.
Gemini CLI Google
Supported
“Call Google Gemini via the @google/gemini-cli npm package in OpenSandbox.”
Source github.com
First-party documented runnable integration example; not behaviorally exercised during this scan.
OpenCode opencode
Supported
“Run the OpenCode coding agent inside an OpenSandbox container.”
Source github.com
First-party documented runnable integration example; not behaviorally exercised during this scan.
Qwen Code Alibaba (Qwen)
Supported
“Run Qwen Code inside an OpenSandbox container through an OpenAI-compatible endpoint.”
Source github.com
First-party documented runnable integration example; not behaviorally exercised during this scan.
GitHub Copilot GitHub
Supported
“Run the official GitHub Copilot CLI inside an OpenSandbox container.”
Source github.com
First-party documented runnable integration example; not behaviorally exercised during this scan.
OpenClaw Peter Steinberger
Supported
“Launch an OpenClaw Gateway inside an OpenSandbox instance and expose its HTTP endpoint.”
Source github.com
First-party documented runnable integration example; not behaviorally exercised during this scan.
Kimi cli
Supported
“Run Kimi CLI (Moonshot AI) in a sandbox”
Source github.com
Advertised example; unregistered harness slug surfaced for separate admission. No registry changes made.
Cursor Anysphere
Supported
“| `cursor` | `./.cursor/rules/` or `~/.cursor/rules/` |”
Source github.com
Advertised agent-skill installation target, not a claim of a sandbox-hosted IDE. Cursor also has an explicit MCP integration.
Windsurf Codeium
Supported
“| `windsurf` | `./.windsurfrules` or `~/.windsurfrules` |”
Source github.com
Advertised agent-skill installation target, not a claim of a sandbox-hosted IDE.
Cline Cline
Supported
“| `cline` | `./.clinerules` or `~/.clinerules` |”
Source github.com
Advertised agent-skill installation target, not a claim of a sandbox-hosted IDE.

Plausible compatibility 0

Derived, not stored: the product passes SSH access, so any CLI coding agent is installable inside it. This is plausible compatibility, not a supported integration.

No general shell access documented, so nothing is derived. Absence is not unsupported.

05 Compliance & regions Standards held, with ordinal rigor; never a score

Certifications held 0

Checked landing/README, SECURITY.md, license, governance and architecture/security guides. No named held compliance certification/attestation found. CNCF Landscape and OpenSSF Best Practices badges are project ecosystem/process signals, not SOC 2/ISO/HIPAA/GDPR attestations.

Regions —

No region list captured.

06 Open source Apache-2.0 · fetched by machine, not quoted from marketing

Licence & repository

Licence: Apache-2.0OSI
Primary language: Python
Stars: 15,803
Forks: 1,455
Created: 17 Dec 2025
Archived: No

Activity

Last commit: 12 Oct 2026
Latest release: release-1.1.1 · 9 Oct 2026
Issues open / closed: 66 / 455 · 87% closed

Recent platform tags observed: release-1.1.0 (2026-09-21), release-1.1.1-rc.1 (2026-09-28), release-1.1.1 (2026-10-09). SDK/component releases are tagged separately; cadence is an observation, not an OSS-health score.

Fetched via gh on 12 Oct 2026.

Self-hosting

Methods: dockerkuberneteshelmsource-buildpackage-manager
You operate: dockerpython

Software is free under Apache-2.0; the operator pays for hardware/cloud, storage, networking, administration and external model usage. Local execution requires Docker and Python 3.10+. Cluster deployment requires Kubernetes/Helm; secure runtimes require installation and compatible nodes/KVM. FastSandbox additionally requires its external fast-sandbox platform, artifact object storage and node-state setup; container pause snapshots need an OCI registry. PostgreSQL is optional for eligible multi-active metadata deployments and Redis is optional for distributed client pools/advanced integration. The operator provides DNS/TLS, authenticated browser access, secret protection and retention. Capability availability is backend/version-dependent; no first-party hosted service is presented.

Documentation signals 7/8

SupportedQuickstart
SupportedAPI reference
SupportedSelf-host guide
SupportedChangelog
SupportedOpenAPI
Not supportedllms.txt
SupportedExamples
SupportedArchitecture doc

07 Company & presence Facts only, not scored

Company

Headquarters: Not disclosed
Founded: —
Team size: Not disclosed
Hiring: unknown

Notable people

jwx0925
hittyt
Pangjiping
ninan-nn

Activity

Last release: 1.1.1 github.com
Last blog post: — open-sandbox.ai

Presence 4

Links found on the public surface scanned; not an exhaustive search.

08 Sources & method How this entry was produced

Public surface only

Everything here was read from open-sandbox.ai's unauthenticated pages, docs, pricing and repository on 12 Oct 2026. Nothing behind a login was used.

Statuses, not scores

Capabilities are pass, partial or fail against a written rubric (features v21). “Not assessed” means the scan predates the key.

Confidence is separate

Each fact carries its own confidence. Low-confidence facts are shown, never hidden, so you can weigh them.

Schema: scan v5
Features registry: v21
Keys tracked today: 61
Assessed on this entry: 61

Something wrong on this page?

Nearby in the atlas

All Sandbox products →