Code & AI agent sandboxes: read the guide

nono

Open source toolSandbox

Ephemeral micro sandboxes for AI agents, with zero setup and zero latency.

Homepage Docs Repository

Scanned 7 Aug 2026 · Schema v5 · features v17

Licence: Apache-2.0
Self-hosted; you pay for the infrastructure.
Free tier: Free to run
Isolation: Not disclosed
the public docs do not say
Regions: —
Not disclosed on the public surface.
Capabilities: 8 / 54
9 partial · 12 fail · 7 not assessed
Agent readiness: 9.5/10 · 3.5/7
read · drive, pre-login surface

Advertised for

AI coding agents · Agent tool invocations (git/gh/curl/kubectl/package managers) · MCP servers · Untrusted scripts (python/node/go) · Ci jobs · Llm/ml workloads

Built for

Developers, then technical

· onboarding is CLI-first and nothing else: 'curl -fsSL https://nono.sh/install.sh | sh' / 'brew install nono' is the entire hero
· hero content is a terminal transcript (`nono search pi` → `nono run --profile nolabs-ai/pi -- pi`), not a product screenshot
· vocabulary is kernel-level throughout — Landlock LSM, seccomp-notify, SECCOMP_IOCTL_NOTIF_ADDFD, Seatbelt, cgroup v2, TOCTOU, SPIFFE SVID

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

Isolation & lifecycle

Isolation primitive

Not disclosedthe public docs do not say

Persistence

Host filesystem — the sandbox operates directly on your real working directory (no volume mounts, no copy-in/copy-out). Sandbox lifetime = the command's lifetime; session state, audit records and rollback snapshots persist under $XDG_STATE_HOME/nono (0700/0600).

Snapshots / forks

Atomic rollback snapshots of the workspace (content-addressable, baseline + final) with `--rollback`; `nono rollback` supports summary, unified/side-by-side diff, restore-to-baseline or to a snapshot, --dry-run preview, and retention pruning. NOT a fork of a running instance.

Resources & limits

Storage

Path-level allow/deny over the host filesystem (--allow / --read / --allow-file, deny-by-default), with a built-in sensitive-path blocklist that wins even inside an allowed parent (~/.ssh, ~/.aws, ~/.gnupg, ~/.kube, ~/.docker, ~/.npmrc, ~/.netrc, ~/.git-credentials, shell history and rc files; on macOS a always-active deny_keychains_macos group). No platform-provided storage.

Limits & quotas

Linux only, supervised runs only, fail-closed: --memory sets a cgroup v2 memory.max over the whole process tree (with memory.swap.max=0 and memory.oom.group=1, so the tree is killed together and cannot spill to swap; exit 137 with a diagnostic) and --max-processes sets pids.max (fork bomb bound — the kernel refuses the fork with EAGAIN, it kills nothing). Requires cgroup v2 with a systemd Delegate=yes user session and the relevant delegated controller; nono REFUSES the run rather than leaving a requested limit unenforced, and refuses write access overlapping /sys/fs/cgroup while a limit is active. NO CPU or I/O limits ('for those, reach for a container').

GPU

Access

Deploy interface

clisdk

SDK languages

RustPythonTypeScriptGoc-ffi

Network & secrets

Egress controls

The differentiator. Default-deny outbound via a supervisor-side HTTP proxy: --block-net (no outbound at all), --allow-domain host allowlists with optional per-PATH globs, six preset network profiles (minimal / developer / claude-code / codex / opencode / enterprise) composed from named host groups, L7 method+path endpoint policy per credential route, DNS-rebinding protection (the proxy resolves and checks every resolved IP against a deny list), cloud-metadata endpoints denied even through an enterprise upstream proxy, a 256-bit per-session proxy token, and --upstream-proxy/--upstream-bypass for corporate egress. The child can only reach localhost:<proxy-port>; all other outbound TCP is blocked at the kernel level.

Secrets

Credential brokering, not env-var handoff: secrets live in the OS keystore (macOS Keychain / Linux Secret Service), 1Password, or Apple Passwords, and the child receives a PHANTOM session token — the proxy swaps it for the real key at the HTTP header level en route to the upstream, so the real credential never enters the sandbox and is zeroised on exit. Per-route endpoint policy limits which methods/paths a credential may be used against. SPIFFE/SPIRE JWT-SVID workload identity is supported as an alternative to static keys (direct bearer injection or an OAuth2 jwt-bearer client assertion). Best-effort redaction scrubs secret-bearing argv/headers/query params from session metadata, audit records and attestations.

Bring your own cloud

Placement & use

Jurisdiction

Advertised workloads

AI coding agentsAgent tool invocations (git/gh/curl/kubectl/package managers)MCP serversUntrusted scripts (python/node/go)Ci jobsLlm/ml workloads

02 Capabilities 8 pass · 9 partial · 12 fail · 54 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 0/3

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

Not applicable

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 applicable

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.

Not applicable

GPU computeGPU-backed instances available.

Not applicable

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).

Partially 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 1/3

Automatic HTTPSTLS endpoints provisioned automatically per workload.

Not applicable

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

Not applicable

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 applicable

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 applicable

Developer access 2/4

Public APIDocumented programmatic API.

Not supported

OpenAPI specMachine-readable OpenAPI / Swagger spec published.

Not supported

CLIOfficial command-line tool.

Supported

SDKOfficial client library/libraries.

Supported

SSH accessDirect SSH into workloads.

Not applicable

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

Not applicable

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

Not applicable

Private container registryHosted private image registry.

Not applicable

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 3/9

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.

Not 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).

Supported

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

Partially 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 applicable

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).

Not assessed

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.

Not applicable

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).

Partially 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).

Partially supported

Organisation & security 2/6

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.

Partially supported

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

Partially 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.

kernel-LSM in-process sandbox (Landlock + seccomp-notify / Seatbelt) — FACET VALUE proposal for isolation_primitive, not a featureos-lsm-sandbox
“nono uses Landlock (Linux and Windows/WSL2) and Seatbelt (macOS) to create irrevocable, kernel-enforced allow-lists. No root, no containers, no overhead.” nono.sh
Per-tool-invocation sandbox brokeringper-tool-sandbox-brokering
“Every tool call gets an ephemeral micro tool sandbox, scoped to only the files, network routes, arguments, and credentials it needs. [...] The agent can ask for a tool, but it cannot widen that tool's sandbox, mint new keys, or bypass endpoint policy from inside the session.” nono.sh
Phantom credential injection (real secret never enters the workload)phantom-credential-injection
“The sandbox receives a phantom GH_TOKEN; the real credential stays with the supervisor and never enters the child. The proxy validates the phantom token, evaluates HTTP method and path policy, injects the real credential at the boundary, and forwards over TLS to api.github.com.” nono.sh
L7 endpoint policy (per-method/per-path authorization on outbound API calls)l7-endpoint-policy
“gh api passes the broad argv rule, but the POST to the repository comments endpoint is denied at the proxy; nothing reaches GitHub.” nono.sh
Tamper-evident, Merkle-rooted audit chaintamper-evident-audit-chain
“Security-relevant events are hash-chained into an audit record sealed with a SHA-256 Merkle root [...] the child can generate events by doing things, but it does not control the audit writer.” nono.sh
Human-in-the-loop approval gate on out-of-policy invocationshuman-approval-gate
“If an action falls outside the provided policy, the policy engine can route it to a human approval decision to approve, deny, or time out.” nono.sh
Signed policy-pack registry with Sigstore verification + trusted publishingsigned-policy-pack-registry
“A nono pack is a signed bundle of artifacts — profiles, hooks, plugins, instructions, scripts, trust policies, groups — distributed via the registry and verified with Sigstore at install time.” nono.sh
SPIFFE workload identity instead of static credentialsworkload-identity-spiffe
“nono integrates with the SPIFFE Workload API to authenticate outbound API calls using workload identity. [...] JWT-SVIDs are currently supported.” nono.sh

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

Can an agent read it?

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

9.5/10

9 full · 1 partial · 95% of applicable checks

llms.txt
Supported
“https://nono.sh/docs/llms.txt — 10,296 bytes, '# Nono Docs' with every one of 76 pages linked as a .md URL plus a one-line description each. (Apex /llms.txt returns the Next.js 404 shell — per rubric v3 the docs-path copy earns full credit.)”
llms-full.txt
Supported
“https://nono.sh/docs/llms-full.txt — 796,388 bytes, all 76 pages concatenated with `Source:` provenance headers. Comprehensive, not truncated. (Retained as a POINTER stub in artifacts/ per the size cap; the pages scored against are saved individually.)”
robots.txt AI policy
Partially supported
“https://nono.sh/robots.txt — 'User-Agent: *\nAllow: /' plus two Sitemap: lines. Default-allow with no AI-UA mention either way → half per the rubric.”
JSON-LD
Supported
“Homepage carries 6 ld+json blocks incl. SoftwareApplication {applicationCategory: SecurityApplication, operatingSystem: 'macOS, Linux, Windows', offers: {price: '0', priceCurrency: 'USD'}, downloadUrl}, plus Organization and WebSite. Rich, product-level. (Organization name is the stale 'Always Further' — see meta.notes.)”
Sitemap
Supported
“https://nono.sh/sitemap.xml — 95 <loc> entries with lastmod/changefreq/priority; a second https://nono.sh/docs/sitemap.xml with 76 more. Both declared in robots.txt.”
Meta tags
Supported
“Descriptive <title> 'Sandbox for AI Agents — Kernel-Level Isolation | nono', a substantive meta description, a full OG set incl. a generated og:image (1200x630), a 14-term keywords list, and rel=canonical.”
Markdown negotiation
Supported
“Every docs page has a clean .md twin — https://nono.sh/docs/cli/clients/claude.md returns markdown prefixed with '> ## Documentation Index / > Fetch the complete documentation index at: https://nono.sh/docs/llms.txt / > Use this file to discover all available pages before exploring further.' An agent is actively pointed at the index.”
Content efficiency
Supported
“Homepage: 8,131 bytes of readable text out of 134,732 bytes of HTML = 6.03% (>5%). Server-rendered Next.js — the full hero, feature copy and testimonials are in the initial HTML.”
Docs discoverable
Supported
“Docs are first-class in the header nav (Sandboxes / Features / Docs), present in the no-JS HTML, and reachable at https://nono.sh/docs without any client-side routing.”
Machine-readable pricing
Not applicable
“No commercial offering to price — a free Apache-2.0 self-hosted tool with no hosted tier. /pricing 404s and the JSON-LD declares price '0' USD. Dropped from the denominator per the oss-tool kind-pack rather than scored as a failure.”
No anti-bot wall
Supported
“No challenge anywhere on the public surface: homepage, all docs pages, .md variants, both sitemaps, robots.txt, install.sh and registry.nono.sh all fetched headless with plain curl/firecrawl at HTTP 200.”

Can an agent drive it?

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

3.5/7

3 full · 1 partial · 50% of applicable checks

Public API
Not supported
“No HTTP/socket control API. The product surface is `nono` (CLI) plus in-process libraries; registry.nono.sh exposes no documented API (/api → HTTP 500, /openapi.json → 404). Scored fail rather than na because a local tool can expose a local API — microsandbox in this dataset does.”
OpenAPI spec
Not supported
“No spec at any probed path on nono.sh or registry.nono.sh; none referenced in the 76-page docs index.”
MCP server
Not supported
“No first-party MCP server. /.well-known/mcp returns the Next.js 404 shell; no npm/pip-installable MCP package is documented. All 11 'MCP' hits in llms-full.txt are about sandboxing OTHER MCP servers (`nono run --block-net --open-port 3000 -- node server.js`) or the tagline 'AI agents, MCP and LLM workloads' — verified against the rubric's mcp_server false-positive guard.”
SDK packages
Supported
“Four official SDKs over one Rust core: nono (crates.io, docs.rs/nono), nono-py, nono-ts, nono-go — each with its own API-reference tree in the docs (/docs/python/api/* 10 pages, /docs/go/api/* 4, /docs/typescript/* 8).”
Auth documented
Not applicable
“No hosted API to authenticate against — nono runs locally and holds no account. (There ARE deep credential docs — keyring, 1Password, OAuth capture, SPIFFE — but those authenticate the WORKLOAD's outbound calls, not a consumer calling nono; registry publishing uses GitHub OIDC trusted publishing, a publisher flow.)”
Self-serve credentials
Not applicable
“There is nothing to sign up for: `brew install nono` (or curl/apt/dnf/AUR/Nix) and run. No account, no key, no card at any point — the check has no object.”
Webhooks
Not supported
“No webhooks. nono has 'hooks' in the agent sense (Claude Code / Codex PostToolUse hook scripts shipped in packs), which is a different mechanism — no outbound HTTP event delivery, signed or otherwise.”
CLI tool
Supported
“Official, first-party, and the primary interface. Homebrew + .deb + Fedora COPR + RPM + AUR (nono-ai-bin, officially maintained) + nixpkgs + a checksum-verifying install.sh; a dedicated 65 KB flag reference page.”
Error model
Partially supported
“Typed SDK errors are documented (https://nono.sh/docs/go/api/errors; Python/TS error surfaces) and several commands emit machine-readable output (`nono audit --json`, `nono rollback --json`, `nono profile show --raw --json`, exit 137 on an OOM kill with a structured diagnostic). But there is no JSON error model in the API sense — there is no API.”
Rate limits documented
Not applicable
“No hosted endpoint to rate-limit. The nearest analogue is local resource capping (--memory / --max-processes), documented thoroughly but a different concept.”
Agent config files
Supported
“Two surfaces: (1) the repo ships AGENTS.md AND CLAUDE.md (plus a NOGENT.md) at its root; (2) packs are a documented distribution channel for agent skills and plugins — `skills/*/SKILL.md`, `.claude-plugin/plugin.json`, `hooks/hooks.json`, with `nono pull` synthesising the Claude Code marketplace wiring. No /.well-known/agent on the web surface (404).”

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

Advertised integrations 10

Claude Code Anthropic
Supported
“Sandboxing Anthropic Claude Code with nono — `nono run --profile nolabs-ai/claude -- claude`”
Source nono.sh
Deepest integration: a registry pack shipping the profile, a Claude Code plugin manifest, hooks/hooks.json, and a skills/nono-sandbox/SKILL.md. Docs advise disabling Claude Code's own sandbox so nono is the single kernel-enforced boundary.
Codex CLI OpenAI
Supported
“Sandboxing OpenAI Codex CLI with nono”
Source nono.sh
Dedicated docs page + `codex` network profile (developer groups + a bundled openai credential) + a PostToolUse hook that surfaces sandbox denials to the agent loop verbatim.
OpenCode opencode
Supported
“Sandboxing OpenCode AI coding assistant with nono”
Source nono.sh
Dedicated docs page + an `opencode` network profile with a bundled google-ai credential.
OpenClaw Peter Steinberger
Supported
“Sandboxing OpenClaw AI coding agent with nono”
Source nono.sh
Dedicated docs page; the networking docs show the gateway invocation (`nono run --profile nolabs-ai/openclaw --proxy-port 19999 --listen-port 18789 -- openclaw gateway`). The launch PR names OpenClaw as the incident that motivated per-tool sandboxing.
GitHub Copilot GitHub
Supported
“GitHub Copilot — https://registry.nono.sh/packages/nolabs-ai/copilot-cli”
Source nono.sh
Named in the landing 'Works with your agents' grid with a registry pack link (nolabs-ai/copilot-cli); no dedicated docs page.
Pi earendil-works
Supported
“$nono search pi ↳ nolabs-ai/pi — $nono run --profile nolabs-ai/pi -- pi”
Source nono.sh
Used as the homepage quickstart example; registry pack nolabs-ai/pi.
Hermes Agent Nous Research
Supported
“Hermes — https://registry.nono.sh/packages/nolabs-ai/hermes”
Source nono.sh
Registry pack link in the landing agent grid.
Goose Block
Supported
“Works with your agents — nono sandboxes any terminal agent. Pull a signed profile from the registry and run — no wrappers, no rewrites. [...] Goose”
Source nono.sh
Named in the landing agent grid but WITHOUT a registry pack link (unlike the others) — advertised compatibility, no shipped pack found at scan time.
Antigravity Google
Supported
“Antigravity — https://registry.nono.sh/packages/nolabs-ai/antigravity”
Source nono.sh
NOT YET IN harnesses.yml — proposed for admission (Google's agentic IDE). Registry pack nolabs-ai/antigravity.
Qwen Code Alibaba (Qwen)
Supported
“Works with your agents [...] Qwen Code”
Source nono.sh
NOT YET IN harnesses.yml — proposed for admission (Alibaba's Qwen Code CLI). Named in the landing grid without a registry pack link.

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.

Legal

Privacy notice

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

Licence & repository

Licence: Apache-2.0OSI
Primary language: Rust
Stars: 3,533
Forks: 236
Created: 31 Jan 2026
Archived: No

Activity

Last commit: 6 Aug 2026
Latest release: v0.71.0 · 31 Jul 2026
Issues open / closed: 109 / 536 · 83% closed

103 tagged releases since 2026-01-31 — roughly weekly minors (v0.65.1 2026-06-23 → v0.71.0 2026-07-31, seven releases in six weeks) plus patch points. Pre-1.0: the README states 'In the lead-up to a 1.0 release, APIs are stabilizing. API changes may still occur where necessary.' Observation, not a graded velocity.

Fetched via gh on 7 Aug 2026.

Self-hosting

Methods: package-managerbinarysource-build
You operate: —

This IS the only way to run it — nono is a local CLI/library, not a service you point at. Zero external dependencies: a single ~2 MB binary, 'no daemon, no container, no VM, and no disk space usage'. Cost = free under Apache-2.0 plus the machine you already own (and your own coding-agent subscription/keys). KERNEL-DEPENDENT: Linux needs Landlock (kernel 5.13+; nono targets ABI v5 and degrades gracefully on older ABIs); macOS uses Seatbelt; Windows is supported ONLY inside WSL2 (84% feature coverage — capability expansion is unavailable because WSL2 lacks seccomp-notify, and proxy-based network filtering is blocked by default), Windows native is NOT supported. Resource limits (--memory / --max-processes) are Linux-only (cgroup v2 + a systemd Delegate=yes user session) and supervised-runs-only.

Documentation signals 7/8

SupportedQuickstart
SupportedAPI reference
SupportedSelf-host guide
SupportedChangelog
Not supportedOpenAPI
Supportedllms.txt
SupportedExamples
SupportedArchitecture doc

07 Company & presence Facts only, not scored

Company

Headquarters: London, United Kingdom
Founded: 2025
Team size: Not disclosed
Hiring: Actively hiring — 2 open engineering roles (Software Engineer; Staff Solutions Engineer), both Remote/UK-preferred, full time

Notable people

Luke Hinds — co-founder & CEO; created Sigstore while at Red Hat
Stephen Parkinson — co-founder
Sal Kimmich — engineering/security (blog byline)

Activity

Last release: v0.71.0 — 2026-07-31 github.com
Last blog post: 2026-07-27 — 'Agent identity, not agent keys' (SPIFFE/SVID workload identity) nono.sh

Presence 9

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 nono.sh's unauthenticated pages, docs, pricing and repository on 7 Aug 2026. Nothing behind a login was used.

Statuses, not scores

Capabilities are pass, partial or fail against a written rubric (features v17). “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: v17
Keys tracked today: 61
Assessed on this entry: 54

Something wrong on this page?

Nearby in the atlas

All Sandbox products →