# NVIDIA OpenShell

The safe, private runtime for fleets of autonomous AI agents.

Source: https://wheredoirunthis.com/products/openshell · JSON: https://wheredoirunthis.com/api/products/openshell · Read on 2026-10-26

## At a glance

- Type: Sandbox
- Kind: Open source tool
- Isolation: Container
- Free tier: Not documented
- Starting price: Not disclosed
- GPU: Not documented
- Regions: Not disclosed
- Coverage: 21 of 61 assessed capabilities supported
- Agent readiness: read 10 of 10 signals, drive 8 of 9 signals

## Links

- Homepage: https://docs.nvidia.com/openshell/about/overview
- Docs: https://docs.nvidia.com/openshell/latest/
- Repository: https://github.com/NVIDIA/OpenShell

## Website design

Desktop public website · assessed 2026-10-26

Complete assessment.

The public entry is a documentation-led site: a white three-column layout, dark sans-serif headings, NVIDIA-green accents and restrained rules. Overview prose and a pale-blue release callout lead the landing; installation uses the same navigation shell with bordered command panels. This assessment covers anonymous desktop landing and documentation views only; pricing is not applicable to the free self-host-only SKU.

- Styles: Corporate, Editorial minimalism
- Landing page: captured
- Pricing: not applicable
- Docs: captured

## Runtime

- **GPU & accelerators**: NVIDIA CDI for Docker/Podman, Kubernetes nvidia.com/gpu scheduling, one GPU for MicroVM; actual hardware supplied by operator. [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/runtimes.md)
- **BYOC (Bring your own cloud)**: na — "No vendor-hosted control plane deploying into customer accounts; this is self-hosted software." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) (confidence: medium)
- **Secrets management**: Provider credentials remain outside the workload; inject only for approved endpoints. Supports OAuth refresh, AWS SigV4, Google auth and SPIFFE JWT-SVID dynamic token grants. Plain --env secrets are readable by the sandbox. [source](https://docs.nvidia.com/openshell/latest/how-it-works/providers/profiles.md)
- **Storage options**: Operator-managed named volumes/PVCs/VM overlays and explicit driver config; gateway metadata uses SQLite locally or PostgreSQL for HA. [source](https://docs.nvidia.com/openshell/latest/how-it-works/gateways/configuration.md)
- **Supported workloads**: autonomous AI agents, coding agents, policy-confined shell/command execution, custom OCI workloads, GPU-backed agent workloads [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md)
- **Persistence**: Driver-supported persistent storage: Docker/Podman named volumes, Kubernetes workspace PVCs, VM disk overlays. Stop/start preserves driver-backed files, not process memory. [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/runtimes.md)
- **Data jurisdiction**: na — "Compute geography and legal jurisdiction depend on the operator's deployment." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) (confidence: medium)
- **Resource limits & quotas**: Operator-configured CPU/memory, GPU counts and gRPC limits; MicroVM ignores --cpu/--memory (use vcpus/mem_mib), VM rootfs archive default 10 GiB, exec input limit 4 MiB; no vendor fleet quota. [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md)
- **SDK languages**: Python, TypeScript, Go, Rust [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md)
- **Network & egress controls**: Default-deny outer network fence; all TCP/DNS traverses a trusted supervisor with executable-aware endpoint/L7 policy checks and endpoint-bound credential injection. [source](https://docs.nvidia.com/openshell/latest/about/architecture.md)
- **Snapshots & forks**: No — "No runtime snapshot/fork operation documented; restart replaces supervisor/main process and does not restore process memory or terminal history." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md) (confidence: medium)
- **Deployment methods**: cli, sdk, api, dockerfile, container-image, template [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md)
- **Isolation**: Container — "Container-based Docker/Podman/Kubernetes are auto-detected; VM is opt-in and never auto-detected. MicroVM uses Apple Hypervisor on macOS or KVM on Linux, not a disclosed Firecracker implementation. Kubernetes runtime_class_name can select Kata when installed. Linux Landlock/seccomp and a separated supervisor enforce policy across drivers." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/runtimes.md) (confidence: high)

## Capabilities

### AI & agents

| Capability | Status | Evidence |
| --- | --- | --- |
| Agent skills | Supported | "The skills teach your agent to drive the OpenShell CLI, write sandbox policies, and debug gateways and inference routing." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| AI app builder | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Autonomous QA agent | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Built-in coding agent | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| BYOK for LLMs | Supported | "OpenCode detects the OpenRouter credential and starts with the model passed to `-m`." [source](https://docs.nvidia.com/openshell/latest/about/run-your-first-agent.md) |
| Credential-injection proxy | Supported | "Agents never see real credentials; OpenShell adds them only to requests bound for approved endpoints." [source](https://docs.nvidia.com/openshell/latest/about/architecture.md) |
| LLM gateway | Supported | "Grant a sandbox access to model providers through provider profiles, attachments, and native endpoints." [source](https://docs.nvidia.com/openshell/latest/how-it-works/inference.md) |
| MCP client | Not supported | "Network rules mediate MCP requests made by sandbox workloads." [source](https://docs.nvidia.com/openshell/latest/how-it-works/policies/network-rules.md) |
| MCP-server hosting | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| MCP server | Partially supported | "For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/openshell/_mcp/server." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Model inference API (first-party) | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Preinstalled coding agent | Not supported | "The default sandbox image is minimal Ubuntu with no agent installed." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |

### Networking & delivery

| Capability | Status | Evidence |
| --- | --- | --- |
| Automatic HTTPS | Not supported | "Exposed services still use the gateway listener and its TLS configuration in this release." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md) |
| Custom domains | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Direct public IPv6 | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Edge KV store | Not applicable | "OpenShell is an agent sandbox runtime, not a managed database or edge key-value service." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| Egress controls | Supported | "Kernel controls confine which files it can access and which system calls it can make, and every network connection passes through a policy check before it leaves the sandbox." [source](https://docs.nvidia.com/openshell/latest/about/architecture.md) |
| Email send / receive | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Public preview URLs | Partially supported | "Pass an optional service name to create a named service URL:" [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md) |
| Private networking | Not supported | "The outer fence denies everything else, so the agent cannot reach a service, the gateway, DNS, or another private address directly." [source](https://docs.nvidia.com/openshell/latest/about/architecture.md) |
| Built-in reverse proxy | Supported | "Service forwarding makes a long-running process inside a sandbox reachable through a gateway-managed URL." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md) |

### Organisation & security

| Capability | Status | Evidence |
| --- | --- | --- |
| Built-in auth | Supported | "Gateways can validate OpenID Connect access tokens on gRPC requests." [source](https://docs.nvidia.com/openshell/latest/how-it-works/gateways/authentication.md) |
| BYOC | Not applicable | "Self-hostable runtime: deployment infrastructure and geographical placement are operated by the user; no maintained first-party hosted OpenShell tier identified." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| Deploy into customers' VPCs | Not applicable | "Self-hostable runtime: deployment infrastructure and geographical placement are operated by the user; no maintained first-party hosted OpenShell tier identified." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| Secrets management | Supported | "Create and manage credential providers that inject API keys and tokens into OpenShell sandboxes." [source](https://docs.nvidia.com/openshell/latest/how-it-works/providers/overview.md) |
| Self-host / OSS core | Supported | "This project is licensed under the [Apache License 2.0](https://github.com/NVIDIA/OpenShell/blob/main/LICENSE)." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| Sharing | Partially supported | "openshell service expose my-sandbox 8080 web" [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md) |
| SSO / SAML | Supported | "Configure OIDC when you want users or automation to authenticate with an identity provider such as Keycloak, Entra ID, or Okta." [source](https://docs.nvidia.com/openshell/latest/how-it-works/gateways/authentication.md) |
| Teams | Supported | "Create OpenShell workspaces, assign members, and understand platform and workspace roles." [source](https://docs.nvidia.com/openshell/latest/how-it-works/workspaces.md) |
| Workload identity | Supported | "Dynamic token grants that use the sandbox's SPIFFE JWT-SVID as an OAuth2 client assertion and inject short-lived tokens into supported headers for matching profile endpoints." [source](https://docs.nvidia.com/openshell/latest/how-it-works/providers/profiles.md) |

### Developer access

| Capability | Status | Evidence |
| --- | --- | --- |
| CLI | Supported | "curl -LsSf https://raw.githubusercontent.com/NVIDIA/OpenShell/main/install.sh \| sh" [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| E2B-compatible API | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Filesystem API | Not supported | "openshell sandbox upload/download provides file transfer; no dedicated non-SSH filesystem API was found in the SDK/lifecycle documentation." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md) |
| Git deploy | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| HTTP data API | Not applicable | "OpenShell is an agent sandbox runtime, not a managed database or edge key-value service." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| OpenAPI spec | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Private container registry | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Public API | Supported | "The primary gateway listener serves the gRPC reflection v1 protocol." [source](https://docs.nvidia.com/openshell/latest/how-it-works/gateways/authentication.md) |
| SDK | Supported | "SDKs connect applications to an OpenShell gateway. They do not install the CLI. Use the same OpenShell release for the SDK and the gateway when possible." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| SSH access | Supported | "the CLI obtains a new SSH session and reattaches to the same main process." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md) |

### Compute & storage

| Capability | Status | Evidence |
| --- | --- | --- |
| Cloud-init / ignition config | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Desktop / computer-use sandbox | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| GPU compute | Supported | "Kubernetes honors counted `--gpu` requests by setting the `nvidia.com/gpu` limit. Docker and Podman select the requested number of default NVIDIA CDI devices in round-robin order. VM gateways accept only one GPU, either through `--gpu` or `--gpu 1`; a single `gpu_device_ids` entry works with either form." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md) |
| Live-state (memory) snapshots | Not supported | "It does not restore process memory or terminal history." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md) |
| Multi-region | Not applicable | "Self-hostable runtime: deployment infrastructure and geographical placement are operated by the user; no maintained first-party hosted OpenShell tier identified." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| Object storage | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Persistent disk | Supported | "Mount existing named volumes or `tmpfs` through driver config." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/runtimes.md) |
| Sandbox templates / custom images | Supported | "Define reusable sandbox workload settings and create sandboxes from templates." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/templates.md) |
| Scale to zero | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Selectable isolation runtime | Supported | "Each gateway runs agent workloads on one compute runtime, selected by its compute driver. Pick the runtime that matches the isolation boundary and infrastructure you need." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/runtimes.md) |
| Snapshots / forking | Not supported | "It does not restore process memory or terminal history." [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md) |

### Data

| Capability | Status | Evidence |
| --- | --- | --- |
| Database branching | Not applicable | "OpenShell is an agent sandbox runtime, not a managed database or edge key-value service." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| Ephemeral instant DB (no-signup) | Not applicable | "OpenShell is an agent sandbox runtime, not a managed database or edge key-value service." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| First-party database | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Vector search | Not applicable | "OpenShell is an agent sandbox runtime, not a managed database or edge key-value service." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |

### Platform & operations

| Capability | Status | Evidence |
| --- | --- | --- |
| Managed CI/CD runners | Not applicable | "Self-hostable runtime: deployment infrastructure and geographical placement are operated by the user; no maintained first-party hosted OpenShell tier identified." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md) |
| Built-in observability | Supported | "Scrape Prometheus metrics from the OpenShell gateway, including supervisor session, relay capacity, peer routing, and mutation lock signals for multi-replica deployments." [source](https://docs.nvidia.com/openshell/latest/observability/gateway-metrics.md) |
| Preview environments | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Scheduled tasks | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Project template marketplace | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |
| Durable workflows | Not supported | "Not found as a shipped OpenShell capability across the retained current documentation index, linked guide/reference corpus, README, and source metadata." [source](https://docs.nvidia.com/openshell/latest/llms.txt) |

## Agent readiness

### Read: 10 of 10 signals present

10/10 signals (7 full, 3 partial)

- Llms txt: pass — "https://docs.nvidia.com/openshell/latest/llms.txt — substantive index with installation, lifecycle, policies, SDKs and architecture links."
- Llms full txt: partial — "https://docs.nvidia.com/openshell/latest/llms.txt — requested llms-full.txt redirects to llms.txt index, not complete page bodies."
- Robots AI policy: pass — "https://docs.nvidia.com/robots.txt — explicitly names GPTBot/Claude/Perplexity/Google-Extended and allows Markdown and llms.txt; no blanket AI Disallow."
- JSON LD: partial — "https://docs.nvidia.com/openshell/about/overview — BreadcrumbList and TechArticle, not a SoftwareApplication/Product with structured offers."
- Sitemap: pass — "https://docs.nvidia.com/openshell/sitemap.xml — nontrivial product sitemap discovered via robots.txt -> fern-sitemap-index.xml."
- Meta quality: pass — "https://docs.nvidia.com/openshell/about/overview — descriptive title, description, OG title/description/image/URL."
- Markdown negotiation: pass — "https://docs.nvidia.com/openshell/about/overview — Accept: text/markdown returns text/markdown; linked .md pages are readable."
- Content efficiency: partial — "https://docs.nvidia.com/openshell/about/overview — visible text excluding script/style/noscript: 4.74% of HTML bytes (1–5% bracket); artifacts/site-signals.json records the current byte counts."
- Docs discoverable: pass — "https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md — first-class Documentation badge and deep guide links; landing itself is the current guide."
- Pricing machine readable: na — "No first-party maintained OpenShell hosted tier/pricing. Apache-2.0 self-host software; own infrastructure costs are not tiers."
- Anti bot: pass — "https://docs.nvidia.com/openshell/about/overview and linked .md corpus fetched without a public-content challenge; Brev JS launchable rendered successfully after consent."

### Drive: 8 of 9 signals present

8/9 signals (6 full, 2 partial)

- Public API: pass — "https://docs.nvidia.com/openshell/latest/how-it-works/gateways/authentication.md — public openshell.v1.OpenShell gRPC API advertised by reflection; four SDKs are documented in the README. Endpoint is self-hosted."
- Openapi spec: na — "Current API is gRPC/protobuf, not a documented REST/Swagger service; common OpenAPI/Swagger paths 404 and no spec found in source index."
- MCP server: pass — "https://docs.nvidia.com/openshell/latest/llms.txt — documented MCP endpoint https://docs.nvidia.com/openshell/_mcp/server; documentation-only, not sandbox operations."
- SDK packages: pass — "https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md — Python, TypeScript, Go and Rust SDK install references."
- Auth docs: pass — "https://docs.nvidia.com/openshell/latest/how-it-works/gateways/authentication.md — mTLS, OIDC, client credentials, roles/scopes and credential layouts documented."
- Agent self serve creds: na — "No hosted signup/key issuance service; operator owns self-hosted gateway credentials and IdP."
- Webhooks: fail — "https://docs.nvidia.com/openshell/latest/llms.txt — no documented signed webhook integration found in the current corpus; gateway gRPC interceptors are not webhooks."
- CLI tool: pass — "https://docs.nvidia.com/openshell/latest/about/installation.md — official openshell CLI installer/packages documented."
- Error model: partial — "https://docs.nvidia.com/openshell/latest/sdk/api-errors.md — detailed gRPC status/protobuf errors (BadRequest, ErrorInfo, RetryInfo), not the rubric's full-credit JSON error model."
- Rate limit docs: partial — "https://docs.nvidia.com/openshell/latest/how-it-works/gateways/configuration.md — grpc_rate_limit_requests/window_seconds settings; no hosted HTTP rate-limit response-header contract."
- Agent config: pass — "https://raw.githubusercontent.com/NVIDIA/OpenShell/main/AGENTS.md and https://raw.githubusercontent.com/NVIDIA/OpenShell/main/README.md — AGENTS.md plus installable NVIDIA/OpenShell agent skills."

## Coding harnesses

- OpenCode (opencode) — "OpenCode detects the OpenRouter credential and starts with the model passed to `-m`." [source](https://docs.nvidia.com/openshell/latest/about/run-your-first-agent.md)
- Claude Code (Anthropic) — "openshell sandbox create --from registry.example.com/your-org/claude-agent:latest -- claude" [source](https://docs.nvidia.com/openshell/latest/how-it-works/sandboxes/overview.md)
- Codex CLI (OpenAI) — "This example builds a sandbox image with Codex, runs its app server inside an OpenShell sandbox, exposes the WebSocket service through the local gateway, and connects a Codex client running on the host." [source](https://raw.githubusercontent.com/NVIDIA/OpenShell/main/examples/codex-app-server/README.md)
- Pi (earendil-works) — "Build a Pi coding-agent image, configure OpenRouter access, and run Pi inside an OpenShell sandbox." [source](https://docs.nvidia.com/openshell/latest/tutorials/run-pi-with-openrouter.md)

Plausible compatibility (derived, not stored):

- Aider
- Amp
- Antigravity
- Devin
- Droid
- Gemini CLI
- Goose
- Hermes Agent
- Kilo Code
- Kiro CLI
- Qwen Code

## Open source

- Repository: https://github.com/NVIDIA/OpenShell
- License: Apache-2.0
- Stars: 19954
- Forks: 1972
- Self-host methods: docker, compose, kubernetes, helm, binary, package-manager
- Self-host dependencies: docker, podman, kubernetes, agent-sandbox, postgres

## Company & presence

- Privacy policy: https://www.nvidia.com/en-us/about-nvidia/privacy-policy/
- Terms of service: https://docs.nvidia.com/openshell/latest/resources/license.md
- GitHub: https://github.com/NVIDIA/OpenShell
- GitHub Discussions: https://github.com/NVIDIA/OpenShell/discussions
- Documentation: https://docs.nvidia.com/openshell/latest/
- PyPI: https://pypi.org/project/openshell/
- NVIDIA Blog (vendor-wide): https://blogs.nvidia.com/?s=OpenShell

## Nearby in the atlas

- [Agent Safehouse](/products/agent-safehouse.md)
- [Agent Sandbox](/products/agent-sandbox.md)
- [Agent Substrate](/products/agent-substrate.md)
- [Agent37](/products/agent37.md)
- [AgentNest](/products/agentnest.md)
- [Anthropic Sandbox Runtime (srt)](/products/sandbox-runtime.md)

## Method

Facts come from the product's own public pages. Each claim carries its evidence: a quote and the page it came from. We count capabilities rather than grade products, and "Not assessed" means we have not checked, never that the product lacks it.