Code & AI agent sandboxes: read the guide

AX (Agent Executor)

Open source toolSandbox

Declare an agentic task. AX runs it at scale.

Homepage Docs Repository

Scanned 25 Sep 2026 · Schema v5 · features v18

Licence: Apache-2.0
Self-hosted; you pay for the infrastructure.
Free tier: Free to run
Isolation: Kata / gVisor
hardened or user-space kernel
Regions: —
Not disclosed on the public surface.
Capabilities: 8 / 55
14 partial · 20 fail · 6 not assessed
Agent readiness: 4/10 · 2.5/9
read · drive, pre-login surface

Advertised for

Interactive coding agents · Long running agent servers · Jupyter notebooks · Headless browser testing · Custom tool runtimes · RL trajectory collection and large scale agent evals

Built for

Developers, then technical

· hero is a terminal session: `ax apply -f task.yaml`, `ax ssh`, `ax suspend`
· vocabulary is Kubernetes, atespaces, actors, gVisor, Redis Streams
· onboarding requires a Kubernetes cluster, Agent Substrate, Go, ko and a registry

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

Isolation & lifecycle

Isolation primitive

Kata / gVisorhardened or user-space kernel

Persistence

Per-task state (memory + filesystem) is checkpointed by Agent Substrate to object storage on suspend and restored on resume; it lasts until `ax delete`. AX control-plane state lives in Redis.

Snapshots / forks

Checkpoint/restore via `ax suspend` / `ax resume` (sub-second resume claimed). Forking a task is roadmap.

Resources & limits

Storage

Workspaces are materialized inside the sandbox at /workspace/<name> (git clones, MCP config, skills). No attachable volumes documented in AX.

Limits & quotas

Per-task CPU/memory requests and limits. Workspace goal bootstrap times out after 10 minutes by default (AX_BOOTSTRAP_TIMEOUT). Token/timeout budgets are roadmap. Scale ceiling is the operator's cluster ('billions of tasks per cluster' is a design claim).

GPU

none documented

Access

Deploy interface

cliapi

SDK languages

Network & secrets

Egress controls

Inherited from Agent Substrate: all actor TCP egress except DNS goes through an egress policy enforcement point (HTTP(S) only; WebSocket, CONNECT, other TCP and UDP blocked), with hostname/CIDR policies configured on Substrate. AX's own Gateway allowlist primitive was removed 2026-09-24.

Secrets

Model API keys come from Kubernetes Secrets referenced by the Model resource. Task env is literal key/value. Credential injection at egress is a Substrate capability, no longer configurable through AX.

Bring your own cloud

Placement & use

Jurisdiction

Advertised workloads

Interactive coding agentsLong running agent serversJupyter notebooksHeadless browser testingCustom tool runtimesRL trajectory collection and large scale agent evals

02 Capabilities 8 pass · 14 partial · 20 fail · 55 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 2/8

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).

Partially 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.

Not assessed

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.

Not supported

Multi-regionCompute placeable in more than one geographic region.

Partially supported

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 applicable

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

Not 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.

Not assessed

Networking & delivery 0/5

Automatic HTTPSTLS endpoints provisioned automatically per workload.

Not applicable

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

Partially supported

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

Not applicable

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

Not applicable

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).

Partially 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 applicable

Developer access 2/8

Public APIDocumented programmatic API.

Supported

OpenAPI specMachine-readable OpenAPI / Swagger spec published.

Not supported

CLIOfficial command-line tool.

Supported

SDKOfficial client library/libraries.

Partially supported

SSH accessDirect SSH into workloads.

Partially supported

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

Partially 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 assessed

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 applicable

AI & agents 3/11

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

Partially 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.

Not supported

MCP clientActs as an MCP client / consumes MCP tools.

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).

Partially 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).

Partially 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).

Not assessed

Data 0/0

First-party databasePlatform-provided managed database.

Not applicable

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 applicable

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 applicable

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 applicable

Platform & operations 0/4

Scheduled tasksCron / scheduled job primitive.

Not supported

Durable workflowsDurable execution / long-running workflow primitive.

Partially 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 applicable

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).

Not 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 applicable

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 1/6

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

Not supported

SSO / SAMLEnterprise single sign-on.

Not supported

Secrets managementFirst-class secrets storage / injection.

Partially supported

TeamsMulti-user team accounts.

Not supported

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

Not 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 assessed

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 assessed

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.

Goal-driven generative workspace setup: describe the ready environment in plain language and a bundled agent installs toolchains/dependencies before the task startsgenerative-environment-setup
“A binding can also carry a goal, a plain-language description of the environment the task needs. On first boot the runner hands that goal to an agent that finishes the setup” github.com
Declarative workspace bundles: git repos + MCP servers/registries + skill registries declared once and materialized into every sandbox bound to themdeclarative-workspaces
“Declare it once, bind it from as many tasks as you like, and the runner materializes it inside each sandbox before the command starts” github.com
Actor multiplexing: many suspended stateful tasks share a small pool of pre-warmed workers, woken on requestactor-multiplexing
“Dozens of tasks share worker resources, turning idle waiting time into spare compute capacity” agentexecutor.io

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

Can an agent read it?

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

4/10

2 full · 4 partial · 40% of applicable checks

llms.txt
Not supported
“https://agentexecutor.io/llms.txt → 404; /docs/llms.txt → 404”
llms-full.txt
Not supported
“https://agentexecutor.io/llms-full.txt → 404”
robots.txt AI policy
Partially supported
“robots.txt: 'User-agent: * / Disallow: /demo/'. Silent default-allow, no AI UA mention.”
JSON-LD
Not supported
“No <script type="application/ld+json"> in homepage HTML.”
Sitemap
Partially supported
“sitemap.xml lists 11 URLs, but every <loc> is relative (e.g. '/docs/quickstart/'), which is invalid per the sitemap protocol. Not referenced from robots.txt.”
Meta tags
Partially supported
“Description + og:title/og:description present and descriptive; <title> is just 'AX'.”
Markdown negotiation
Not supported
“Accept: text/markdown returns text/html; /docs/quickstart/index.md → 404.”
Content efficiency
Supported
“~5.7 KB readable text / 20.3 KB HTML ≈ 28%, server-rendered (Hugo 0.140.0).”
Docs discoverable
Partially supported
“Docs exist at /docs/ (sitemap), but the homepage links only to GitHub ('Get started' → github.com/google/ax#quick-start); no /docs link in the homepage HTML.”
Machine-readable pricing
Not applicable
“No pricing: free Apache-2.0 self-host tool with no commercial tier.”
No anti-bot wall
Supported
“Fastly-cached static pages; no challenge on any public path fetched.”

Can an agent drive it?

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

2.5/9

2 full · 1 partial · 28% of applicable checks

Public API
Supported
“DESIGN.md '## API reference': ax.v1alpha1.AX gRPC service, Task/Workspace/Model CRUD + Suspend/Resume/Watch.”
OpenAPI spec
Not supported
“No OpenAPI. pkg/apis/v1alpha1/ax.proto is published (gRPC), which the rubric does not credit as OpenAPI.”
MCP server
Not supported
“No MCP control server. MCP appears only as something workspaces wire INTO sandboxes (mcp-client).”
SDK packages
Partially supported
“Go only: generated types in pkg/apis/v1alpha1 + embeddable runner package.”
Auth documented
Not supported
“No API auth documented; the CLI reaches the control plane through the kube context / tunnel.”
Self-serve credentials
Not applicable
“Self-hosted: there is no vendor-issued credential to obtain.”
Webhooks
Not supported
“No webhooks. WatchTask is a server-streaming RPC, not outbound webhooks.”
CLI tool
Supported
“`go install github.com/google/ax/cmd/ax@latest`; official kubectl-shaped CLI.”
Error model
Not supported
“No documented error model beyond task conditions (e.g. ActorCreationFailed).”
Rate limits documented
Not applicable
“Self-hosted API with no vendor-imposed rate limits.”
Agent config files
Not supported
“No AGENTS.md in the repo, no /.well-known/agent, no skills listing.”

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

Advertised integrations 0

The public surface names no coding harness as supported or documented-to-run-here.

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

No attestations found on the public surface. Empty means nothing is advertised, after checking, not that nothing was checked.

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: Go
Stars: 11,435
Forks: 551
Created: 30 Mar 2026
Archived: No

Activity

Last commit: 25 Sep 2026
Latest release: v0.3.1 · 25 Sep 2026
Issues open / closed: 39 / 161 · 81% closed

7 releases, v0.1.0 (2026-05-20, launch eve) to v0.3.1 (2026-09-25): a minor every ~2 months with patch bursts (v0.2.0-v0.2.2 within two days). GitHub release bodies are empty; no changelog.

Fetched via gh on 25 Sep 2026.

Self-hosting

Methods: kubernetessource-build
You operate: kubernetesagent-substrateredisobject-storage

Self-host is the only way to run it: free under Apache-2.0 + your own Kubernetes cluster. Agent Substrate (pre-1.0, its own install into ate-system) must be running first; object storage for Substrate snapshots (GCS in its GKE quickstart). The control plane is built and deployed from source with `make deploy AX_IMAGE_REPO=<your-registry>` via ko, so you need Go, ko and a pullable registry; no Helm chart. Redis is deployed alongside as the state store. Goal-driven workspace setup needs a GEMINI_API_KEY. v1alpha1: breaking changes expected.

Documentation signals 5/8

SupportedQuickstart
SupportedAPI reference
SupportedSelf-host guide
Not supportedChangelog
Not supportedOpenAPI
Not supportedllms.txt
SupportedExamples
SupportedArchitecture doc

07 Company & presence Facts only, not scored

Company

Headquarters: Mountain View, CA, USA (Google)
Founded: 2026
Team size: Not disclosed
Hiring: Unknown

Notable people

Jaana Dogan (rakyll): lead author (477 commits) and co-author of the launch post
Ethan Bao: Engineering Director, co-author of the launch post

Activity

Last release: v0.3.1 — 2026-09-25 github.com
Last blog post: 2026-05-21 — 'Introducing Agent Executor, Google's distributed Agent Runtime' cloud.google.com

Presence 5

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 agentexecutor.io's unauthenticated pages, docs, pricing and repository on 25 Sep 2026. Nothing behind a login was used.

Statuses, not scores

Capabilities are pass, partial or fail against a written rubric (features v18). “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: v18
Keys tracked today: 61
Assessed on this entry: 55

Something wrong on this page?

Nearby in the atlas

All Sandbox products →