{
 "features_resolved": {
  "snapshot_any": true,
  "snapshot_mem": true,
  "fork": true,
  "pause_resume": true,
  "persistent_disk": true,
  "whole_disk": true,
  "persist_scope": null,
  "volumes": false,
  "auto_stop_idle": true,
  "long_sessions": true,
  "custom_image": true,
  "image_snapshot": true,
  "image_oci": true,
  "vm_isolation": true,
  "docker_inside": true,
  "nested_virt": false,
  "root": true,
  "systemd": null,
  "browser": false,
  "desktop": false,
  "computer_use": false,
  "code_interpreter": true,
  "browser_control": null,
  "desktop_control": null,
  "browser_and_desktop": false,
  "anti_bot": false,
  "captcha_solving": false,
  "residential_ip": false,
  "agent_harnesses": false,
  "ssh": false,
  "public_ipv4": false,
  "inbound_https": true,
  "custom_domain": true,
  "raw_tcp_inbound": false,
  "egress_allowlist": false,
  "open_internet": null,
  "static_egress_ip": false,
  "private_network": null,
  "self_host": false,
  "byoc": false,
  "open_source": false,
  "soc2": true,
  "hipaa": false,
  "sso": true,
  "gpu": false,
  "arm64": false,
  "windows": false,
  "macos": false,
  "lang_python": true,
  "lang_node": true,
  "lang_go": true,
  "lang_rust": true,
  "lang_java": null,
  "gpu_desktop": false,
  "credential_injection": false,
  "egress_http_rules": false,
  "wake_on_request": true,
  "live_resize": true,
  "fork_running": true,
  "memory_snapshot_fork": true,
  "webhooks": false,
  "scheduled_wakeups": null,
  "audit_logs": null,
  "eu_data_residency": false,
  "zero_data_retention": null,
  "scoped_api_keys": null,
  "spend_limits": null,
  "mcp_server": false,
  "usage_api": false,
  "snapshot_auto": true,
  "snapshot_on_demand": true,
  "harness_api": false,
  "own_agent_api": false,
  "fast_boot": false,
  "strong_isolation": true,
  "ingress_rules": true,
  "guest_firewall": true,
  "secret_proxy": false,
  "secret_proxy_any": false,
  "volume_attach": false,
  "volume_shared": false,
  "sidecar": null,
  "fuse": true,
  "kernel_features": true,
  "burst": null
 },
 "company": {
  "id": "codesandbox-sdk",
  "checked": "2026-09-30",
  "company": "CodeSandbox B.V.",
  "ownership": {
   "value": "subsidiary",
   "parent": "Together AI",
   "source": "https://codesandbox.io/blog/joining-together-ai-introducing-codesandbox-sdk",
   "quote": "Today, I'm excited to announce that CodeSandbox has joined Together AI!"
  },
  "stage": {
   "value": "subsidiary",
   "source": "https://codesandbox.io/blog/joining-together-ai-introducing-codesandbox-sdk",
   "quote": "Today, I'm excited to announce that CodeSandbox has joined Together AI!",
   "confidence": "high",
   "note": "CodeSandbox was acquired by / merged into Together AI in Dec 2024 (announced by founder Ives van Hoorne as 'joined Together AI', widely reported as an acquisition, e.g. HN thread titled 'CodeSandbox Acquired by Together AI'). Prior to that it had raised a priced seed and Series A as an independent startup; those are recorded below for history."
  },
  "accelerators": [],
  "total_funding_usd": {
   "value": 15100000,
   "status": "confirmed",
   "basis": "sum of confirmed seed ($2.4M, 2019) + Series A ($12.7M, 2020) per CodeSandbox's own blog posts; no further priced rounds reported before the Dec 2024 Together AI acquisition (deal terms undisclosed)",
   "source": "https://codesandbox.io/post/codesandbox-series-a-funding",
   "quote": "we've raised a new round of $12.7 million Series A funding, led by EQT Ventures"
  },
  "last_round": {
   "type": "series-a",
   "amount_usd": 12700000,
   "date": "2020-10",
   "leads": [
    "EQT Ventures"
   ],
   "status": "confirmed",
   "source": "https://codesandbox.io/post/codesandbox-series-a-funding",
   "quote": "we've raised a new round of $12.7 million Series A funding, led by EQT Ventures. With participation from existing investors, Kleiner Perkins, Arches Capital, and Vercel founder Guillermo Rauch accompanied by new angels Andreas Blixt and Daniel Gebler."
  },
  "investors": [
   "Kleiner Perkins",
   "EQT Ventures",
   "Arches Capital",
   "Guillermo Rauch (Vercel founder)",
   "Dylan Field (Figma co-founder)",
   "Quinn Slack (Sourcegraph co-founder)",
   "Christian Bach & Mathias Biilmann (Netlify founders)",
   "Koen Bok (Framer co-founder)"
  ],
  "portfolio_listings": [],
  "revenue": {
   "value": null,
   "status": "unknown"
  },
  "employees": {
   "value": null,
   "source": ""
  },
  "note": "CodeSandbox (founded 2017 by Ives van Hoorne, Amsterdam/Netherlands) was acquired by / merged into Together AI in December 2024; the CodeSandbox product and team continue operating (per the announcement, 'CodeSandbox will continue to run as normal'), and CodeSandbox SDK (this card's product) launched as part of that deal to provide sandboxed code execution for Together AI's generative-AI customers. Funding figures above (seed + Series A, totaling $15.1M) predate the acquisition and were raised while CodeSandbox was an independent startup; acquisition price was not disclosed. Buyers should treat CodeSandbox SDK's backing as that of Together AI (a well-funded AI cloud/inference company), not as a standalone startup's balance sheet."
 },
 "id": "codesandbox-sdk",
 "name": "CodeSandbox SDK",
 "url": "https://codesandbox.io/docs/sdk/pricing",
 "category": "agent-sandbox",
 "status": "deprecated: succeeded by Together's together-sandbox SDK; still sold self-serve through CodeSandbox accounts and on Together custom plans",
 "isolation": "firecracker",
 "modes": [
  {
   "key": "on-demand",
   "label": "VM credits, fixed sizes ($0.01486/credit, per started minute)",
   "features": {
    "isolation": "firecracker"
   },
   "flags": [],
   "os": [
    "linux"
   ],
   "gpu_only": false,
   "pricing": "sizes",
   "vcpu_h": null,
   "ram_gib_h": null,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": 2,
   "max_vcpu": 64,
   "max_ram_gib": 128,
   "sizes": [
    {
     "name": "Pico",
     "vcpu": 2,
     "ram_gib": 1,
     "disk_gib": 20,
     "hour": 0.0743,
     "month_cap": null
    },
    {
     "name": "Nano",
     "vcpu": 2,
     "ram_gib": 4,
     "disk_gib": 20,
     "hour": 0.1486,
     "month_cap": null
    },
    {
     "name": "Micro",
     "vcpu": 4,
     "ram_gib": 8,
     "disk_gib": 20,
     "hour": 0.2972,
     "month_cap": null
    },
    {
     "name": "Small",
     "vcpu": 8,
     "ram_gib": 16,
     "disk_gib": 20,
     "hour": 0.5944,
     "month_cap": null
    },
    {
     "name": "Medium",
     "vcpu": 16,
     "ram_gib": 32,
     "disk_gib": 20,
     "hour": 1.1888,
     "month_cap": null
    },
    {
     "name": "Large",
     "vcpu": 32,
     "ram_gib": 64,
     "disk_gib": 20,
     "hour": 2.3776,
     "month_cap": null
    },
    {
     "name": "XLarge",
     "vcpu": 64,
     "ram_gib": 128,
     "disk_gib": 20,
     "hour": 4.7552,
     "month_cap": null
    }
   ],
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": null,
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "start_fee": 0,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": null,
   "note": "Credits/h: Pico 5, Nano 10, Micro 20, Small 40, Medium 80, Large 160, XLarge 320 at $0.01486/credit (docs; marketing page rounds to $0.015 → Micro $0.30/h). Per started minute (3m25s bills 4 min). Billed on allocated tier, wall-clock while running incl. idle tail until hibernationTimeoutSeconds; hibernated = no credits. Plan caps: Build/Pro max Micro; Scale max Medium (16c/32GB); Large/XLarge Enterprise only. SDK vmTier param goes up to Small unless a custom template is built."
  },
  {
   "key": "enterprise-bulk",
   "label": "Enterprise bulk VM credit packs (up to 50% off)",
   "features": {
    "isolation": "firecracker"
   },
   "flags": [
    "sales"
   ],
   "os": [
    "linux"
   ],
   "gpu_only": false,
   "pricing": "sizes",
   "vcpu_h": null,
   "ram_gib_h": null,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": 2,
   "max_vcpu": 64,
   "max_ram_gib": 128,
   "sizes": [
    {
     "name": "Pico",
     "vcpu": 2,
     "ram_gib": 1,
     "disk_gib": 20,
     "hour": 0.03715,
     "month_cap": null
    },
    {
     "name": "Nano",
     "vcpu": 2,
     "ram_gib": 4,
     "disk_gib": 20,
     "hour": 0.0743,
     "month_cap": null
    },
    {
     "name": "Micro",
     "vcpu": 4,
     "ram_gib": 8,
     "disk_gib": 20,
     "hour": 0.1486,
     "month_cap": null
    },
    {
     "name": "Small",
     "vcpu": 8,
     "ram_gib": 16,
     "disk_gib": 20,
     "hour": 0.2972,
     "month_cap": null
    },
    {
     "name": "Medium",
     "vcpu": 16,
     "ram_gib": 32,
     "disk_gib": 20,
     "hour": 0.5944,
     "month_cap": null
    },
    {
     "name": "Large",
     "vcpu": 32,
     "ram_gib": 64,
     "disk_gib": 20,
     "hour": 1.1888,
     "month_cap": null
    },
    {
     "name": "XLarge",
     "vcpu": 64,
     "ram_gib": 128,
     "disk_gib": 20,
     "hour": 2.3776,
     "month_cap": null
    }
   ],
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": "Prepaid bulk credit packs; pack size and exact discount negotiated. 50% is the published MAXIMUM ('up to 50% off'), so these rates are a best case.",
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "start_fee": 0,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": null,
   "note": "Enterprise plan only (Together sales). Custom concurrency, VMs up to 64 vCPU/128 GiB, optional dedicated cluster and SSO."
  }
 ],
 "storage": {
  "included_disk_gib": 20,
  "disk_gib_month": null,
  "disk_billed_when_stopped": false,
  "snapshot_gib_month": 0,
  "paused_ram_billed": false
 },
 "network": {
  "egress_free_gib": null,
  "egress_gib": null,
  "ipv4_month": null,
  "ipv4_available": false
 },
 "plans": [
  {
   "name": "Build (free)",
   "fee": 0,
   "flags": [],
   "fee_is_credit": false,
   "included_usd": 5.944,
   "per_seat": 0,
   "concurrency": 10,
   "max_session_h": null,
   "max_vcpu": 4,
   "max_ram_gib": 8,
   "sizes_allowed": [
    "Pico",
    "Nano",
    "Micro"
   ],
   "note": "400 credits/month (no rollover). NO on-demand credits: VMs freeze when the 400 credits run out, so total usage is capped at $5.94/month. 20 new SDK sandboxes/h, 1,000 SDK requests/h, 5 members. Verifier 2026-09-28: trial_only so the engine never auto-picks it: Build has no on-demand credits (VMs freeze at 400 credits), so fee 0 + usage - $5.94 would understate every workload above $5.94/month.",
   "trial_only": true
  },
  {
   "name": "Pro (legacy editor plan)",
   "fee": 12,
   "flags": [],
   "fee_is_credit": false,
   "included_usd": 14.86,
   "per_seat": 0,
   "concurrency": 10,
   "max_session_h": null,
   "max_vcpu": 16,
   "max_ram_gib": 32,
   "sizes_allowed": [
    "Pico",
    "Nano",
    "Micro",
    "Small",
    "Medium"
   ],
   "note": "1,000 base credits/month; add-on credit packs with unpublished bulk savings; mandatory spending limit (VMs freeze at the limit). Fee not shown on current pricing page. 20 new sandboxes/h, 1,000 req/h, 20 members."
  },
  {
   "name": "Scale",
   "fee": 170,
   "flags": [],
   "fee_is_credit": false,
   "included_usd": 23.78,
   "per_seat": 0,
   "concurrency": 250,
   "max_session_h": null,
   "max_vcpu": 16,
   "max_ram_gib": 32,
   "sizes_allowed": [
    "Pico",
    "Nano",
    "Micro",
    "Small",
    "Medium"
   ],
   "note": "$170/month per workspace platform fee (not credit). Included credits: pricing page '160 hours of monthly VM credits' = 1,600 credits = $23.78 if Nano-hours; SDK docs example says 1,100 credits ($16.35). On-demand credits $0.01486 billed at month end. 1,000 new sandboxes/h, 10,000 req/h, up to 20 members."
  },
  {
   "name": "Enterprise",
   "fee": null,
   "flags": [
    "sales"
   ],
   "fee_is_credit": null,
   "included_usd": null,
   "per_seat": 0,
   "concurrency": null,
   "max_session_h": null,
   "max_vcpu": 64,
   "max_ram_gib": 128,
   "sizes_allowed": null,
   "note": "Contact Together sales: bespoke concurrency and request limits, up to 50% off bulk credit packs, unlimited members, dedicated support."
  }
 ],
 "extras": [
  {
   "item": "SSO (Enterprise optional extra)",
   "price": null
  },
  {
   "item": "Dedicated cluster (Enterprise optional extra)",
   "price": null
  },
  {
   "item": "Extra VM storage beyond 20 GB ('more on demand', paid plans)",
   "price": null
  },
  {
   "item": "Education / open-source / non-profit discounts",
   "price": null
  }
 ],
 "free": {
  "monthly_credit": 5.944,
  "one_time_credit": 0
 },
 "features": {
  "linux": true,
  "windows": false,
  "macos": false,
  "arm64": false,
  "gpu": false,
  "gpu_types": [],
  "snapshot": "mem",
  "pause_resume": true,
  "fork": true,
  "persistent_disk": true,
  "volumes": false,
  "auto_stop_idle": true,
  "max_session_h": null,
  "custom_image": true,
  "docker_inside": true,
  "nested_virt": false,
  "root": true,
  "systemd": null,
  "browser": false,
  "desktop": false,
  "computer_use": false,
  "code_interpreter": true,
  "agent_harnesses": false,
  "ssh": false,
  "public_ipv4": false,
  "inbound_https": true,
  "custom_domain": true,
  "raw_tcp_inbound": false,
  "egress_allowlist": false,
  "static_egress_ip": false,
  "private_network": null,
  "self_host": false,
  "byoc": false,
  "open_source": false,
  "regions": null,
  "soc2": true,
  "hipaa": false,
  "sso": true,
  "lang_python": true,
  "lang_node": true,
  "lang_go": true,
  "lang_rust": true,
  "lang_java": null,
  "gpu_desktop": false,
  "credential_injection": false,
  "egress_http_rules": false,
  "wake_on_request": true,
  "live_resize": true,
  "fork_running": true,
  "memory_snapshot_fork": true,
  "webhooks": false,
  "scheduled_wakeups": null,
  "audit_logs": null,
  "eu_data_residency": false,
  "zero_data_retention": null,
  "scoped_api_keys": null,
  "spend_limits": null,
  "mcp_server": false,
  "isolation": "firecracker",
  "image_oci": true,
  "snapshot_auto": true,
  "snapshot_on_demand": true,
  "ingress_rules": true,
  "secret_proxy": false,
  "anti_bot": false,
  "captcha_solving": false,
  "volume_attach": false,
  "volume_shared": false,
  "harness_api": false,
  "own_agent_api": false,
  "max_session_unlimited": true,
  "egress_ip_type": "datacenter",
  "fast_boot": false,
  "usage_api": false
 },
 "perf": {
  "cpu_runs_s": null,
  "cold_start_ms": 2329.37,
  "burst_tti_ms": 7657.92,
  "dax_s": 81.23
 },
 "perf_source": "ComputeSDK (github.com/computesdk/benchmarks results/): cold 2,329.37 ms (2026-08-07, SDK default tier), burst 7,657.92 ms (2026-09-25, composite 14.4), dax 81.23 s from the 2026-08-28 run on VMTier.Small (8 vCPU / 16 GiB); later DAX runs failed. Same VMs for on-demand and enterprise-bulk.",
 "knobs": [
  {
   "id": "hibernation_tail",
   "label": "Idle tail before hibernation (x session hours)",
   "kind": "slider",
   "min": 1,
   "max": 3,
   "step": 0.05,
   "default": 1,
   "unit": "x",
   "options": null,
   "effect": "hours_factor",
   "modes": [
    "on-demand",
    "enterprise-bulk"
   ],
   "why": "VMs bill full credits while running idle until hibernationTimeoutSeconds fires: SDK default 300 s on free workspaces, 1800 s on paid ones (max 86400 s); hibernated VMs consume no credits, and HTTP/WebSocket traffic auto-wakes them. Factor = (session + timeout) / session, e.g. the paid 30-min default on 30-min sessions = 2x, 5 min on 15-min sessions = 1.33x. Default 1 assumes you hibernate explicitly or set a short timeout.",
   "source": "https://github.com/codesandbox/codesandbox-sdk/blob/main/src/types.ts"
  },
  {
   "id": "resume_rounding",
   "label": "Extra per-minute rounding from resume cycles",
   "kind": "slider",
   "min": 1,
   "max": 1.5,
   "step": 0.05,
   "default": 1,
   "unit": "x",
   "options": null,
   "effect": "hours_factor",
   "modes": [
    "on-demand",
    "enterprise-bulk"
   ],
   "why": "Billing is per started minute (3m25s bills 4 min) and each resume starts a new run. The card rounds once per session; with k hibernate/resume runs per session add ~0.5 min average rounding per extra run: factor ~ 1 + (k-1)*0.5/D minutes (judgment: uniform remainder). Max 1.5 = e.g. 11 runs in 10-min sessions.",
   "source": "https://codesandbox.io/docs/sdk/pricing"
  }
 ],
 "caveats": [
  "Credit price conflict: docs/FAQ $0.01486 vs pricing page $0.015 (0.94% higher); card uses $0.01486.",
  "Scale included credits conflict: pricing page '160 hours' (read as 1,600 credits = $23.78) vs SDK docs example 1,100 credits ($16.35) — the docs example is stale (also says 100 concurrent VMs; now 250).",
  "free.monthly_credit ($5.94) is the Build plan allowance (400 credits, no rollover). Build is a hard cap (no on-demand credits; VMs freeze), so the Build plan is marked trial_only and never auto-picked; workloads under $5.94/month are really free but the estimate prices them on Scale.",
  "Per-started-minute billing (min 60 s per run); short resume/exec/hibernate cycles are heavily rounded up.",
  "Pico spec conflict: SDK docs 2 cores/1 GB vs pricing page 1 core/2 GB.",
  "Hibernated/shut-down sandboxes consume no credits and no storage fee is documented (snapshot_gib_month 0 assumed from the documented 2-component billing model: credits + plan). Hibernated sandboxes are archived after ~7 days (resume 20-60 s).",
  "Idle tail billed at full rate until hibernationTimeoutSeconds; SDK default is 300 s on free and 1800 s on paid workspaces (src/types.ts: 'Defaults to 300 seconds for free users, 1800 seconds for pro users. Maximum is 86400 seconds (1 day).', checked 2026-10-14). Scale/Pro VMs left on the default bill 30 idle minutes after every session unless hibernate() is called or the timeout lowered; the hibernation_tail knob defaults to 1x. Automatic wake-up on HTTP/WebSocket restarts billing.",
  "Egress not metered or published (null). No public IPv4.",
  "Enterprise-bulk mode uses the maximum published discount (50%); actual discount negotiated. Pro add-on pack discounts unpublished; legacy Pro fee unpublished.",
  "Regions: no selectable or guaranteed region. StartSandboxOpts.ipcountry is a best-effort scheduling hint (src/types.ts: 'Country, served as a hint on where you want the sandbox to be scheduled. For example, if \"NL\" is given as a country, the sandbox will be scheduled in a cluster inside Europe. Note that this is not a guarantee'), so a European cluster exists but features.regions stays null and eu_data_residency false. Checked 2026-10-14.",
  "perf from benchmarks.json providers.codesandbox (computesdk): cold start median 2,329 ms, burst 7,658 ms, dax 81.23 s (dax last run success rate 0; value from 2026-08-28 run). No hpc CPU benchmark.",
  "Verified 2026-09-28 by rendering https://codesandbox.io/pricing, /docs/sdk/pricing, /docs/learn/credit-usage/credits, /docs/learn/plans/subscriptions, /docs/learn/plans/pricing-faq in headless Chrome: size table, $0.01486 (docs) vs $0.015 (page), per-started-minute billing, 400 / 1,000 subscription credits, Scale $170 + 160 h credits + $0.15/h on-demand, 10/10/250 concurrency, Build 4/8, Pro/Scale 16/32, Enterprise 64/128 and up to 50% off packs all confirmed.",
  "Re-checked 2026-09-28 (docs/sdk/faq and docs/sdk/pricing via headless Chrome): no SSH access and no egress/bandwidth price documented for SDK VMs (access is via the SDK Terminals/Commands API). ssh stays null, egress null.",
  "Resume from an archived sandbox takes 10-60 s (resume docs) and archived forks 20-60 s (create docs); sandboxes are archived 'after period of inactivity' (no day count on current pages; the old ~7 days figure could not be re-verified: /docs/sdk/hibernate now shows the SDK intro).",
  "Pro and Scale fees are shown as 'From $12' / 'From $170' per month per workspace; the monthly-vs-annual split behind 'From' is not published on the page (annual plans exist per the Mar 2024 changelog). Card uses the 'From' price.",
  "Deprecated product (checked 2026-10-14, https://docs.together.ai/docs/together-code-sandbox-legacy): 'This page covers the deprecated code sandbox SDK. Use the together-sandbox SDK for new integrations.' / 'The together-sandbox SDK is the successor to this offering.' / 'The legacy code sandbox is available on Together's custom plans.' / 'A self-serve option is available by creating an account with CodeSandbox.' Still sold at the same $0.01486/credit, so it stays priced. Last SDK release 2.4.2 (2025-12-04, CHANGELOG). The successor (https://docs.together.ai/docs/together-code-sandbox) is a separate product and not covered by this card.",
  "codesandbox.io pages were behind a Cloudflare bot check on 2026-10-14 (headless Chrome and WebFetch got 'Just a moment...'); that day's checks used Together docs and the SDK GitHub repo."
 ],
 "sources": [
  "https://codesandbox.io/pricing",
  "https://codesandbox.io/docs/sdk/pricing",
  "https://codesandbox.io/docs/learn/credit-usage/credits",
  "https://codesandbox.io/docs/learn/credit-usage/controlling-usage",
  "https://codesandbox.io/docs/learn/plans/subscriptions",
  "https://codesandbox.io/docs/learn/plans/usage-based-billing",
  "https://codesandbox.io/docs/learn/plans/pricing-faq",
  "https://codesandbox.io/docs/sdk/faq",
  "https://codesandbox.io/docs/sdk/create",
  "https://codesandbox.io/docs/sdk/resume",
  "https://codesandbox.io/docs/sdk/lifecycle",
  "https://codesandbox.io/docs/sdk/specs",
  "https://docs.together.ai/docs/together-code-sandbox-legacy",
  "https://github.com/codesandbox/codesandbox-sdk/blob/main/src/types.ts"
 ],
 "fleet": {
  "default_max_instances": 10,
  "default_max_vcpu": null,
  "default_max_gpus": null,
  "raise": "self-serve",
  "raise_time_days": null,
  "verification": null,
  "tiers": "Build free 10; Scale 250; Enterprise custom via Together sales.",
  "notes": "Chrome-rendered SDK pricing explicitly lists 10/250/custom; a stale worked example lower on the same page says Scale 100. Current tier table used; 300/1,000 need sales. Credits/runtime budgets are separate from concurrency.",
  "sources": [
   "https://codesandbox.io/docs/sdk/pricing"
  ],
  "researched": "2026-09-28"
 },
 "audit_2026_09_29": {
  "applied": 4,
  "source": "checked against the provider's live pricing and docs pages",
  "reviewed": 2
 },
 "feature_evidence": {
  "snapshot_auto": {
   "value": true,
   "source": "https://codesandbox.io/docs/sdk/core-concepts",
   "quote": "By default, when a Sandbox is hibernated, a memory snapshot is created. This allows for near-instant resumes the next time the Sandbox is accessed—typically within 0.5–2 seconds.",
   "note": "Triggered by the idle timeout set at create time (https://codesandbox.io/docs/sdk/create: 'hibernationTimeoutSeconds ... How quickly the sandbox should hibernate after inactivity'). Memory + disk + running processes included (resume docs: 'restoring it to its exact previous state including file system, memory, and running processes'). Automatic wake-up on HTTP/WebSocket resumes it. Old snapshots can expire/archive, then boot is CLEAN (disk only, setup tasks rerun). shutdown() skips the snapshot.",
   "checked": "2026-09-29"
  },
  "snapshot_on_demand": {
   "value": true,
   "source": "https://github.com/codesandbox/codesandbox-sdk/blob/main/README.md",
   "quote": "Snapshotting/restoring VMs (checkpointing) at any point in time",
   "note": "Implemented as hibernate() (POST /vm/{id}/hibernate), which stops the VM and saves a memory+disk snapshot ('Snapshots: Disk/Memory snapshot that is generated when hibernating a Sandbox and consumed when resuming a Sandbox', key-terms). New sandboxes can boot from it: create({id: <sandbox-id>}) forks a hibernated sandbox in 1-3 s; templates are memory/fs snapshots. Live forks of a RUNNING sandbox (max 5, share memory with parent). One current snapshot per sandbox (no list of named restore points documented); take a checkpoint without stopping = live fork.",
   "checked": "2026-09-29"
  },
  "ingress_rules": {
   "value": true,
   "source": "https://codesandbox.io/docs/sdk/create",
   "quote": "private: Sandbox is private, hosts require tokens",
   "note": "Per-sandbox privacy: public / private / public-hosts (default public-hosts). Private hosts need signed host tokens (URL, headers or cookies; https://codesandbox.io/docs/sdk/hosts: 'access your Sandbox hosts securely using signed urls, headers or cookies'). Preview tokens can be created/revoked (/sandbox/{id}/tokens). HTTPS preview hosts only; no raw TCP, no IP allowlists documented.",
   "checked": "2026-09-29"
  },
  "secret_proxy": {
   "value": false,
   "source": "https://github.com/codesandbox/codesandbox-sdk/blob/main/CHANGELOG.md",
   "quote": "support env and cwd in shell api",
   "note": "Secrets reach the sandbox only as environment variables passed to commands/terminals; no credential-injecting proxy documented.",
   "checked": "2026-09-29"
  },
  "nested_virt": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/create",
   "quote": "Every Sandbox is backed by a Firecracker VM and is completely isolated and persisted",
   "note": "Not documented as available. Sandboxes are Firecracker microVMs; Firecracker does not expose hardware virtualization extensions (/dev/kvm) to guests, and no SDK/docs page mentions nested virtualization or running VMs/emulators inside a Sandbox. Derived from the isolation technology, not an explicit statement.",
   "checked": "2026-09-30"
  },
  "browser": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/faq",
   "quote": "Does the CodeSandbox SDK use VM or Browser Sandboxes? ... the actual environments that it uses run in a VM.",
   "note": "not offered: outside the product's scope. The SDK provisions general-purpose Linux dev/code-exec VMs (terminals, commands, file system, hosts); there is no managed/real browser instance shipped inside the sandbox or offered as a service. 'Browser' in the docs refers only to connecting to the sandbox FROM a browser client, not a browser running inside it.",
   "checked": "2026-09-30"
  },
  "anti_bot": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk",
   "quote": "CodeSandbox SDK enables you to programmatically spin up development environments and run untrusted code.",
   "note": "not offered: outside the product's scope. This is a code-execution/dev-VM sandbox product, not a browser-automation or web-scraping product, so stealth/anti-fingerprinting features are not applicable and are not documented anywhere in the SDK docs.",
   "checked": "2026-09-30"
  },
  "captcha_solving": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk",
   "quote": "CodeSandbox SDK enables you to programmatically spin up development environments and run untrusted code.",
   "note": "not offered: outside the product's scope. No CAPTCHA-solving capability is documented; the product is a general dev/code-exec VM sandbox, not a scraping/automation tool.",
   "checked": "2026-09-30"
  },
  "ssh": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/clients",
   "quote": "There are three different ways to connect to a Sandbox: SDK ... Node ... Browser",
   "note": "No SSH client or SSH access is documented anywhere in the SDK docs. All interaction is via the SDK connection (SDK/Node/Browser clients) using Terminals/Commands/File System APIs, not SSH.",
   "checked": "2026-09-30"
  },
  "volume_attach": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/core-concepts",
   "quote": "Sandboxes are isolated virtual machines (powered by Firecracker) ... Persistent file storage across sessions",
   "note": "not offered: outside the product's scope. Each Sandbox has one built-in persistent workspace disk (/project/workspace); no docs mention attaching additional/extra persistent volumes to a Sandbox.",
   "checked": "2026-09-30"
  },
  "volume_shared": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/core-concepts",
   "quote": "Sandboxes are isolated virtual machines (powered by Firecracker) ... Persistent file storage across sessions",
   "note": "not offered: outside the product's scope. There is no attachable-volume feature at all (see volume_attach), so a single volume mounted by multiple running Sandboxes simultaneously is not offered/documented.",
   "checked": "2026-09-30"
  },
  "harness_api": {
   "value": false,
   "source": "https://github.com/codesandbox/codesandbox-sdk/blob/main/README.md",
   "quote": "AI Agents: Create a sandbox for each AI agent, and run the agent in the sandbox.",
   "note": "not offered: outside the product's scope. The SDK lets you run any process (including a coding agent) inside a Sandbox via Terminals/Commands, but there is no first-party API that specifically drives a coding harness (prompt/stream/follow-up); no agent-harness templates or dedicated API are documented.",
   "checked": "2026-09-30"
  },
  "own_agent_api": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk",
   "quote": "CodeSandbox SDK enables you to programmatically spin up development environments and run untrusted code.",
   "note": "not offered: outside the product's scope. CodeSandbox does not ship its own browser/computer-use/research task agent; the SDK is purely sandbox infrastructure for running your own code/agents.",
   "checked": "2026-09-30"
  },
  "mcp_server": {
   "value": false,
   "source": "https://github.com/codesandbox",
   "quote": "(GitHub org repository listing for codesandbox contains no MCP server package)",
   "note": "No MCP server repo or documentation found in the codesandbox GitHub org or SDK docs; not offered as far as could be determined.",
   "checked": "2026-09-30"
  },
  "egress_http_rules": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk",
   "quote": null,
   "note": "not offered: outside the product's scope. No egress control of any kind is documented (see egress_allowlist), so HTTP method/path-level egress rules are not offered.",
   "checked": "2026-09-30"
  },
  "static_egress_ip": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk",
   "quote": null,
   "note": "No static/fixed outbound IP is documented anywhere in the SDK docs or pricing pages; no public IPv4 is offered either (features.public_ipv4=false on the card), and no add-on for a fixed egress IP is mentioned.",
   "checked": "2026-09-30"
  },
  "open_source": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/faq",
   "quote": "Is it possible to self-host CodeSandbox SDK? No, we do not offer self hosting at this time.",
   "note": "The core product (CodeSandbox platform / SDK backend infra) is closed-source/proprietary SaaS; self-hosting is explicitly not offered. The client SDK package on GitHub (codesandbox/codesandbox-sdk) carries no OSI license (license: null via GitHub API), so even the client library isn't clearly open source.",
   "checked": "2026-09-30"
  },
  "eu_data_residency": {
   "value": false,
   "source": "https://codesandbox.io/legal",
   "quote": "CodeSandbox may process personal data in countries within the European Economic Area (“EEA”). Transfer to countries outside of the EEA is also permitted, provided that the provisions of the GDPR are complied with.",
   "note": "No guaranteed region: the SDK only takes a best-effort ipcountry scheduling hint (src/types.ts: 'Note that this is not a guarantee, and the sandbox might end up in a different region based on availability and scheduling decisions'), and the Data Processing Agreement explicitly permits transferring data outside the EEA, so there is no guaranteed EU-only residency option for customers. Re-checked 2026-10-14.",
   "checked": "2026-09-30"
  },
  "max_session_h": {
   "value": "none",
   "source": "https://codesandbox.io/docs/sdk/create",
   "quote": "hibernationTimeoutSeconds: How quickly the sandbox should hibernate after inactivity. With active lifecycle management this is recommended to be set to 24 hours ... Set maximum hibernation timeout of 86400 and do manual hibernation",
   "note": "There is no documented maximum wall-clock lifetime for a running Sandbox; the only automatic stop mechanism is the idle hibernation timeout (configurable; SDK default 300 s free / 1800 s paid, maximum 86400 s per src/types.ts), which is about inactivity, not a hard session-length cap. A continuously-used (non-idle) Sandbox can in principle keep running indefinitely, so treated as 'none' (no hard max session length), distinct from the idle-hibernation behavior.",
   "checked": "2026-09-30"
  },
  "desktop": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk",
   "quote": "The CodeSandbox SDK has been designed to let you to rapidly create and run secure code environments at scale. The SDK is already supporting use cases such as: Browser based IDEs, Code interpretation, AI Agent runtime environments, A/B testing",
   "note": "not offered: outside the product's scope. The SDK is a programmatic dev-VM/code-interpreter API (terminals, commands, file system, HTTPS host previews); no VNC/RDP graphical desktop is documented anywhere in the SDK docs or template library.",
   "checked": "2026-09-30"
  },
  "computer_use": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk",
   "quote": "The SDK is already supporting use cases such as: Browser based IDEs, Code interpretation, AI Agent runtime environments, A/B testing",
   "note": "not offered: outside the product's scope. No screenshot/click driving API; SDK exposes Terminals/Commands/FileSystem/Ports/Hosts clients only, no screenshot+click computer-use API. The GitHub sandbox-templates repo has no desktop/computer-use template (only dev-framework and 'headless-chromium' templates).",
   "checked": "2026-09-30"
  },
  "volumes": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/specs",
   "quote": "Whenever you start a sandbox, you can specify which VM specs to use. ... SDK Sandbox VMTier parameters can be used to create/update environments up to Small (8 Cores/16GB RAM) specs.",
   "note": "Storage is a fixed per-tier VM disk (20 GB across all tiers per the card's sizes table) that persists via hibernate/resume; no separate attachable/extra persistent volume product is documented anywhere in the SDK docs (create/specs/hosts/faq).",
   "checked": "2026-09-30"
  },
  "public_ipv4": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/hosts",
   "quote": "Hosts are the services exposed by your Sandbox. They all have their own hostname which includes the port. The client.hosts API will allow you to access your Sandbox hosts securely using signed urls, headers or cookies.",
   "note": "Access to exposed ports is via per-port HTTPS hostnames/preview URLs (optionally with signed host tokens), not a raw IPv4 address on the VM. No public IPv4 is documented anywhere in the SDK docs.",
   "checked": "2026-09-30"
  },
  "agent_harnesses": {
   "value": false,
   "source": "https://github.com/codesandbox/sandbox-templates",
   "quote": "angular, assemblyline, astro-starlight, astro-tailwind, astro, bun, cloudflare-worker, codeium, create-t3-app, deno-fresh, deno, docker, docusaurus, elixir, gatsby, gleam, go, headless-chromium, hono-next, html-css, javascript, jupyter, nest, nextjs, node-express-server, node-http-server, node, nodebox-starter, nuxt-todos-edge, nuxt, php, postgres-drizzle-nextjs, python-fasthtml, python-flask-server, python-openai, python-pytorch, python-selenium, python-tensorflow, python, qwik-vite, rails, react-*, rust-*, solid-*, storybook-react, sveltekit, tanstack-start, universal, vite*, vue-vite",
   "note": "Official template library lists only framework/runtime/dev-tool templates (one 'codeium' AI-coding-assistant template) — no Claude Code, Codex, OpenCode or similar coding-agent-harness templates preinstalled/offered.",
   "checked": "2026-09-30"
  },
  "gpu": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/specs",
   "quote": "SDK Sandbox VMTier parameters can be used to create/update environments up to Small (8 Cores/16GB RAM) specs. If higher tiers are specified only the core count will increase unless you create a custom template using our CLI.",
   "note": "not offered: outside the product's scope. VM tiers (Pico..XLarge) are CPU/RAM-only Firecracker VMs; no GPU tier, GPU flag, or GPU pricing is documented anywhere in the SDK/pricing docs.",
   "checked": "2026-09-30"
  },
  "windows": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/create",
   "quote": "Sandboxes are the main building blocks of your SDK experience. Every Sandbox is backed by a Firecracker VM and is completely isolated and persisted, providing you with a secure environment to modify, test and run your code.",
   "note": "not offered: outside the product's scope. Firecracker only supports Linux guests; templates are built from Linux devcontainers/Dockerfiles. No Windows machine option is documented.",
   "checked": "2026-09-30"
  },
  "macos": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/create",
   "quote": "Sandboxes are the main building blocks of your SDK experience. Every Sandbox is backed by a Firecracker VM and is completely isolated and persisted, providing you with a secure environment to modify, test and run your code.",
   "note": "not offered: outside the product's scope. Firecracker microVMs are Linux-only; no macOS/Apple-VM offering exists in this product.",
   "checked": "2026-09-30"
  },
  "arm64": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/specs",
   "quote": "SDK Sandbox VMTier parameters can be used to create/update environments up to Small (8 Cores/16GB RAM) specs. If higher tiers are specified only the core count will increase unless you create a custom template using our CLI.",
   "note": "VM tiers are described purely by vCPU/RAM (Pico..XLarge), with no architecture choice documented anywhere in the SDK docs; no ARM/aarch64 option is offered or mentioned.",
   "checked": "2026-09-30"
  },
  "self_host": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/faq",
   "quote": "Is it possible to self-host CodeSandbox SDK? No, we do not offer self hosting at this time.",
   "note": "Directly confirmed in the official SDK FAQ.",
   "checked": "2026-09-30"
  },
  "byoc": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/faq",
   "quote": "Is it possible to self-host CodeSandbox SDK? No, we do not offer self hosting at this time.",
   "note": "The FAQ's blanket 'no self hosting at this time' covers running the service in the customer's own infra/cloud account (BYOC); no separate BYOC/dedicated-cloud offering is documented, and Enterprise features (dedicated cluster, SSO, bulk credits) described elsewhere are add-ons within CodeSandbox's own cloud, not deployment into the customer's account.",
   "checked": "2026-09-30"
  },
  "egress_ip_type": {
   "value": "datacenter",
   "source": "https://codesandbox.io/docs/sdk/specs",
   "quote": "Whenever you start a sandbox, you can specify which VM specs to use.",
   "note": "derived: Sandboxes are Firecracker microVMs running on CodeSandbox's own cloud infrastructure (no residential/mobile network offering documented anywhere in the SDK docs or pricing pages); outbound traffic is not documented as coming from anything other than standard datacenter IP space. No static/fixed egress IP is offered (features.static_egress_ip=false) and no public IPv4 is offered (features.public_ipv4=false), so this is inferred from the infra type rather than an explicit statement.",
   "checked": "2026-09-30"
  },
  "egress_allowlist": {
   "value": false,
   "source": "https://codesandbox.io/docs/sdk/create",
   "quote": "Every Sandbox is backed by a Firecracker VM and is completely isolated and persisted, providing you with a secure environment to modify, test and run your code.",
   "note": "not offered: no outbound/egress control feature (allow list, deny list, no-internet mode, or security-group egress rules) is documented anywhere in the SDK docs (create, hosts, ports, specs, faq) or pricing pages; the only network-facing controls documented are inbound host privacy settings (public/private/public-hosts) for exposed ports.",
   "checked": "2026-09-30"
  },
  "hipaa": {
   "value": false,
   "source": "https://codesandbox.io/pricing",
   "quote": "SOC 2 Type II compliance ... SSO ... Your own dedicated cluster ... Dedicated support",
   "note": "not offered: the pricing page's Enterprise feature list explicitly enumerates its compliance/security add-ons (SOC 2 Type II, SSO, dedicated cluster, dedicated support) and does not mention HIPAA eligibility or a BAA anywhere on this page, the legal/terms page, or the SDK docs, so HIPAA support is not offered as far as could be determined from official pages.",
   "checked": "2026-09-30"
  },
  "fast_boot": {
   "value": false,
   "source": "https://www.computesdk.com/benchmarks",
   "quote": "burst of 100: p95 9843 ms, 100% started",
   "note": "ComputeSDK sandbox benchmark 2026-09-25",
   "checked": "2026-09-25"
  },
  "usage_api": {
   "value": false,
   "source": "https://github.com/codesandbox/codesandbox-sdk",
   "quote": "The SDK exports endpoints for sandboxList, sandboxCreate, sandboxGet, sandboxFork, vmListRunningVms, vmHibernate, vmShutdown, vmStart and other VM management functions, but no usage, billing, or cost retrieval endpoints exist. SDK documentation and API reference contain only sandbox/VM lifecycle man",
   "note": "granularity: account-only; cost: no; CodeSandbox SDK provides programmatic access to create, manage, and control sandboxes, but does not expose REST API endpoints or SDK methods to retrieve billable usage (running time, vCPU-seconds, GB-seconds, or credits consumed) per sandbox, per tag, or per project. Billing/usage information is only available through the CodeSandbox web dashboard and account interface.",
   "confidence": "high",
   "checked": "2026-01-31"
  }
 },
 "locations": [],
 "locations_source": null,
 "locations_quote": "Sandboxes are the main building blocks of your SDK experience. Every Sandbox is backed by a Firecracker VM and is completely isolated and persisted.",
 "locations_checked": "2026-10-01"
}