Cloudflare Sandbox SDK by Cloudflare

# Fix list: Cloudflare Sandbox SDK

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/cloudflare-sandbox-sdk, the October 2026 research run, assessed 1 October 2026. Grade B, 67.8 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on Cloudflare Sandbox SDK: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Payments & pricing, 30 out of 100, up to 8.8 more on the total

Why it scored 30: No x402 or other machine payment on anything the SDK exposes (0). Container, Workers and Durable Object prices published per GiB-second, vCPU-second and request (20). Needs Workers Paid at $5 a month, so no card-free route (0). Stripe Projects lists Cloudflare, which lets an agent create an account through the operator's Stripe login (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 2. Security & auth, 65 out of 100, up to 6.1 more on the total

Why it scored 65: There's no hosted credential to score. Deploys use Cloudflare API tokens, but the sandbox sits behind a Worker you write, the starter template has no auth, and the docs say sandbox IDs aren't cryptographically secure (15). Each sandbox is a separate VM with its own filesystem, process space and network stack (10). Outbound traffic can be cut off with `enableInternet = false` or limited with a deny-by-default `allowedHosts` list and `deniedHosts`, all GA, though internet access is on by default (10). Outbound handlers run in the Worker, outside the container, and inject credentials the container never sees, and S3 mounts can use a credential proxy (15). The 0.12.4 release added sandbox labels for analytics, and we didn't check account audit logs this run (5). security.txt lists HackerOne and a disclosure policy (10). Certifications and public advisories for the SDK weren't checked (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 3. Agent ergonomics, 63 out of 100, up to 6 more on the total

Why it scored 63: Command output, file reads and process logs come back whole or streamed, with no size controls beyond what your code adds (15). Directory listing options and streaming exec, but little filtering (10). Typed errors with codes such as STALE_PREVIEW_URL (15). The same sandbox ID returns the same sandbox, so a retry lands in one place, though tunnels.get() isn't idempotent under wrangler dev (#874) (15). TypeScript only, and the first deploy needs Docker locally and a Worker you write (8).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 4. Schema & documentation, 77 out of 100, up to 3.7 more on the total

Why it scored 77: No OpenAPI for the library. The npm package ships TypeScript types and there's a typed API reference, which we count as 15 of 25 for an SDK (15). llms.txt and Markdown pages for every doc (10). Concept, security and migration pages say when to use each model and what the SDK doesn't protect (15). Typed option objects, few free-form blobs (12). Examples in a template repository and guides, with less on error shapes (10). Dated changelog, GitHub release notes and a 1.0 migration guide (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 5. Reliability, 87 out of 100, up to 2.6 more on the total

Why it scored 87: Scored on the SDK checklist, since the listing is a library. Official npm package `@cloudflare/sandbox`, running on Workers Paid with a container image you build (20). Public GitHub Actions with release, CodeQL and scheduled performance-test runs, all passing on main when checked (25). 23 open issues, several of them data-loss bugs opened in August and September 2026. Backups that silently drop top-level directories (#859), restores of archives of 10 MB or more that can't be recovered (#884), and `allowedHosts` failing closed for approved hosts (#844). We couldn't see replies (12). Release notes flag breaking changes, such as the desktop API removed in 0.12.0 (15). 1.0.0 published on 2026-09-30 (15). Cloudflare's own status page wasn't scored here.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 6. Maintenance & community, 70 out of 100, up to 2.6 more on the total

Why it scored 70: 1.0.0 on npm on 2026-09-30 (30). We found two releases since 3 July 2026, 0.12.4 on 21 July and 1.0.0 on 30 September, and one Sandbox changelog entry, so this line scores nothing (0). 23 open issues and 16 open pull requests, most issues from the last 60 days, reply times not visible (15). It's Cloudflare's official SDK and current (15). Dependabot, CodeQL and CI passing (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## 7. Transparency & trust, 73 out of 100, up to 2.4 more on the total

Made of editorial 45, provenance 100.

Why it scored 73: The SDK and its container image are Apache-2.0. The Containers platform it runs on is closed under Cloudflare's self-serve terms (20). Cloudflare's privacy policy is published, but we didn't read its retention or DPA terms this run, so it scores as a policy with no retention statement found (5). A dated notice that 0.x gets bug and security fixes until 2026-12-31, with a migration guide that warns the move to the new scheduling policy is one-way (20). Container locations and subprocessors weren't checked this run, so not found (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- Whether more releases shipped between 0.12.4 (21 July 2026) and 1.0.0 (30 September 2026). The GitHub releases page didn't show 1.0.0 and we couldn't read the npm version list.
- Whether maintainers have replied to the backup and restore bugs opened in August and September 2026.
- Cloudflare's data-retention, DPA and container-location terms as they apply to Containers, which we didn't read this run.
- Reliability uses the SDK checklist because the listing's kind is sdk. On the hosted-platform checklist it would score well below 87, since we found no Containers rate limits, 429 guidance or SLA for sandboxes.

## Weaknesses

- No hosted API. You deploy and secure a Worker before an agent can call anything, and the starter has no auth
- Open bugs on backups that drop directories (#859) and restores of archives of 10 MB or more (#884)
- Needs Workers Paid at $5 a month, with no card-free route
- TypeScript only
- Two models to learn while 0.x and 1.0 overlap until 31 December 2026

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Derive the sandbox ID from the authenticated user, as the docs advise. IDs aren't secrets
- Put API keys in an outbound handler in the Worker, not in the container's environment
- Set `enableInternet = false` or an `allowedHosts` list before running untrusted code. Internet access is on by default
- Check that a restored backup has every directory you expect. Backups of 10 MB or more have open bugs
- Start new projects on 1.0. The 0.x library only gets fixes until 31 December 2026

## What the review panel asked for

- Fix the backup and restore bugs
- Document Containers rate limits
- auth in starter template

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.