Blaxel Sandboxes by Blaxel

# Fix list: Blaxel Sandboxes

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/blaxel-sandboxes, the October 2026 research run, assessed 1 October 2026. Grade C, 61 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 Blaxel Sandboxes: 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. Reliability, 35 out of 100, up to 13 more on the total

Why it scored 35: Status page at status.blaxel.ai (incident.io) with per-component history and a Sandboxes uptime of 99.48 per cent for July to October 2026 (20). The 90-day record has 25 incidents, and three majors touched sandboxes for over an hour each. Sandbox deploy errors in us-pdx-1 for 3 hours on 13 August, runtime errors in us-pdx-1 for 2 hours 10 minutes on 5 September, and a critical workload outage in us-was-1 for 2 hours 28 minutes on 1 October (0). No request-rate limits found, only concurrency quotas per tier (0). No 429 or Retry-After guidance found, though the error reference marks WORKLOAD_UNAVAILABLE as retryable and tells callers not to retry the rest (5). No SLA found (0). Sandboxes aren't labelled beta (10).

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.

## 2. Payments & pricing, 40 out of 100, up to 7.5 more on the total

Why it scored 40: No x402, MPP or L402. An open issue on the Python SDK asking for x402 dates from 26 September 2026 (0). Per-unit prices published without a login, $0.0000115 a GB of RAM a second (20). Up to $200 of free credits, but we found no statement on whether a card is needed, so half (10). Stripe Projects, since 10 June 2026, lets an agent create the account and receive credentials through the operator's Stripe login and saved card. A person still has to log in to Stripe once (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.

## 3. Security & auth, 64 out of 100, up to 6.3 more on the total

Why it scored 64: API keys that can be set never to expire, OAuth 2.0 client-credentials tokens that last 2 hours, and service accounts limited to one workspace with an admin or member role. We found no per-action scopes (25). Each sandbox is a microVM with no shared kernel, which Blaxel's blog says is Firecracker (10). Domain allow and deny lists with method and path rules, network-level enforcement since 25 September 2026, but the default is open and the docs call domain filtering public preview (7). The proxy can inject secrets into outbound headers so they never enter the sandbox, under the same preview label (12). Process logs per sandbox and traces for a sampled 10 per cent of executions, no audit log found (5). Blaxel claims SOC 2 Type II, ISO 27001 and HIPAA. No security.txt, disclosure policy or bug bounty found, and its compliance portal blocks our fetcher (5).

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.

## 4. Agent ergonomics, 70 out of 100, up to 4.9 more on the total

Why it scored 70: Management lists take `limit` (default 50, maximum 200), `cursor` and `sort`, and the sandbox MCP server has ranged file reads, but there's no field selection (15). Pagination arrived in the SDKs and Management API on 30 June 2026, filtering is thinner (15). Errors carry a code, a status and a retryable flag (20). No idempotency keys, but names conflict with a 409, which makes a retry by name safe. That's worth 10, less 5 because the 18 MCP tools carry no documented read-only or destructive annotations (5). SDKs in TypeScript, Python and Go, and a sandbox needs little more than an image and memory (15).

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.

## 5. Schema & documentation, 80 out of 100, up to 3.3 more on the total

Why it scored 80: OpenAPI 3.0 for the per-sandbox API is public on GitHub, and the control-plane spec (OpenAPI 3.0.3) is rendered in the API reference, though we couldn't download it at the path the docs name (20). llms.txt and Markdown pages (10). Reference entries say what each call does and list response codes, with less on when not to use them (10). Typed fields with required metadata and spec, few enums seen (10). An error-code reference with 11 codes and their HTTP statuses, plus examples in the spec (15). Date-based API versions through the `Blaxel-Version` header and a dated changelog (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.

## 6. Transparency & trust, 68 out of 100, up to 2.8 more on the total

Made of editorial 60, provenance 75.

Why it scored 68: SDKs and sandbox templates are MIT. The platform is closed under dated terms (10 December 2025) (20). A data-collection page lists what error tracking and OpenTelemetry collect, the 10 per cent trace sampling and zero data retention for sandboxes deleted before standby. It gives no retention periods, and the privacy policy sits on a compliance portal our fetcher can't read (15). Date-based API versions with a stated default version, but no deprecation policy with notice periods found (10). Telemetry and opt-outs are documented (`DO_NOT_TRACK`, `TELEMETRY_ENABLED=false`, `BL_ENABLE_OPENTELEMETRY`), and regions are named (us-pdx-1, us-was-1, eu-lon-1, eu-fra-1), but no subprocessor list was readable (15).

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.

Provenance checks not met in full (half of this category, computed from checked facts):

- Domain age: blaxel.ai, no registry record we could read (0 of 15)
- security.txt: not found (0 of 10)

## 7. Maintenance & community, 85 out of 100, up to 1.3 more on the total

Why it scored 85: @blaxel/core 0.3.25 on npm on 2026-09-30 (30). Seven Python SDK releases from 20 July to 28 July and more than ten dated changelog entries since 3 July (20). Three open issues on sdk-python, the oldest from 28 April 2026, and 24 open pull requests on sdk-typescript. We couldn't see reply times (15). Current official SDKs in TypeScript, Python and Go (15). We didn't find CI status for the SDK repositories (5).

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.

## 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 the up-to-$200 credit needs a card. Neither the pricing page nor the getting-started guide says.
- The listing's claim of a `sandboxes:create` scope. We found roles (admin, member) but no per-action scopes, so we've rewritten authNotes without it.
- The privacy policy, DPA and subprocessor list on compliance.blaxel.ai, which disallows automated fetching.
- Why the Python SDK's last PyPI release (0.4.1, 28 July 2026) lags the TypeScript SDK (0.3.25, 30 September 2026).

## Weaknesses

- 25 status-page incidents from 9 July to 1 October 2026, three of them sandbox outages over an hour
- No published request-rate limits, 429 guidance or SLA
- Domain filtering and secret injection are marked public preview, and egress is open by default
- No security.txt, and the privacy policy lives on a portal that blocks automated readers
- Billed by memory, and about half of it goes to the writable tmpfs layer

## 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.

- Close WebSocket and terminal connections when done. An open connection keeps the sandbox active and billed
- Set a TTL or idle expiry on throwaway sandboxes. The default is to keep them
- Set `forbiddenDomains` or an allow list at creation, with `network.firewall` for tools that ignore proxy settings. Both can only be set when the sandbox is created
- Retry only on WORKLOAD_UNAVAILABLE. The error reference says other codes won't succeed on retry
- Connect to `<sandbox URL>/mcp` with your API key instead of wrapping the REST API in tools yourself

## What the review panel asked for

- Publish rate limits and 429 behaviour
- Publish an SLA
- per-sandbox credentials
- GA egress controls

## 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.