Modal Sandboxes by Modal

# Fix list: Modal Sandboxes

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/modal-sandboxes, the October 2026 research run, assessed 1 October 2026. Grade BB, 75.6 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 Modal 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. Payments & pricing, 40 out of 100, up to 7.5 more on the total

Why it scored 40: No x402, MPP or L402 (0). Per-second CPU, memory and GPU prices published (20). Starter has $30 of compute every month and needs no card, per the pricing page (20). Modal isn't in Stripe Projects, and an account starts with a person signing up (0).

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. Agent ergonomics, 67 out of 100, up to 5.4 more on the total

Why it scored 67: Exec output streams, but nothing trims command output or file reads for a context window (15). `Sandbox.list()` filters by tags (15). Typed exceptions, and 1.6.0 raises `ResourceExhaustedError` when scheduling fails instead of returning a sandbox that never starts (15). Named sandboxes are unique per app and a duplicate raises `AlreadyExistsError`, so a retry can't start a second copy, though `from_name()` only finds running ones (10). Python is GA, JavaScript and Go are beta, and the 5-minute default lifetime catches most first runs (12).

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.

## 3. Security & auth, 76 out of 100, up to 4.2 more on the total

Why it scored 76: A token ID and secret pair per workspace, revocable, plus Connect Tokens that open one sandbox's HTTP or WebSocket server to an outside caller (25). gVisor by default, with a full VM runtime on Team and Enterprise (8). Outbound traffic can be blocked or limited to CIDR ranges, GA, with a domain allow list in beta, and no inbound connections without tunnels (10). Modal Secrets go into the sandbox's environment, and we found no proxy that keeps credentials outside it, so this rests on network controls and guidance (8). Audit logs on Enterprise only (10). SOC 2 Type 2, a private HackerOne bug bounty through security@modal.com and stated fix times (24 hours critical, one week high). No security.txt and no public advisories found (15).

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. Schema & documentation, 79 out of 100, up to 3.4 more on the total

Why it scored 79: No REST API or OpenAPI. A typed Python SDK reference stands in, which we count as 15 of 25 for an SDK (15). llms.txt and Markdown pages (10). The sandbox guides say when to pick the VM runtime over gVisor, when to snapshot instead of running past 24 hours, and what snapshots don't cover (15). Typed parameters such as `timeout`, `block_network` and `cidr_allowlist` (12). Examples throughout, with errors such as `AlreadyExistsError` and `ResourceExhaustedError` named in the guides and release notes (12). Versioned release notes for every SDK release (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. Transparency & trust, 74 out of 100, up to 2.3 more on the total

Made of editorial 60, provenance 88.

Why it scored 74: The client SDKs are Apache-2.0. The platform is closed under terms dated May 2026 (20). The security page states retention per product, function inputs and outputs up to 7 days, logs from 1 to 30 or more days by plan, volumes until you delete them, and says Modal won't read code or data without permission (25). Breaking changes are confined to 1.Y.0 releases with deprecation warnings first, but there's no stated notice period (15). Subprocessors and data locations 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.

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

- security.txt: not found (0 of 10)

## 6. Reliability, 95 out of 100, up to 1 more on the total

Why it scored 95: Scored on the SDK checklist, since Modal sandboxes are reached only through its SDKs. Official `modal` package on PyPI with Python 3.10 to 3.14 supported and 3.9 dropped (20). Public GitHub Actions with unit tests, checks, docs and CodeQL, passing on main when checked (25). 17 open issues against more than 9,000 commits (20). Breaking changes go only into 1.Y.0 releases and are called out in the release notes (15). 1.6.0, so past 1.0 (15). For readers, the status page showed one 14-minute dashboard and sandbox incident in mid-September 2026 and nothing else in 90 days.

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.

## 7. Maintenance & community, 93 out of 100, up to 0.6 more on the total

Why it scored 93: 1.6.0 on 2026-09-28 (30). 1.5.4 on 12 August, 1.5.5 on 28 August and 1.6.0 on 28 September (20). 17 open issues on modal-client, response times not visible to us (18). The Python SDK is current, the JavaScript and Go SDKs are beta (15). CodeQL and unit tests running on main (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.

## 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 Modal publishes a subprocessor list and data locations for sandboxes. Not checked this run.
- Whether any workspace token can be limited to sandbox actions only. We found no scoped token type.
- Reliability uses the SDK checklist because the listing's kind is sdk. On the hosted-platform checklist it would score about 50 (status page 20, one 14-minute incident 20, GA 10, and nothing for sandbox rate limits, 429 guidance or an SLA, none of which we found).

## Weaknesses

- No REST API, and the JavaScript and Go SDKs are beta
- Default lifetime of 5 minutes and a hard maximum of 24 hours
- gVisor rather than a VM unless you're on Team or Enterprise for the VM runtime
- Memory snapshots are alpha, kept 7 days, and end the sandbox
- No security.txt, and audit logs only on Enterprise

## 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.

- Pass `timeout=` when you create a sandbox. The default lifetime is 5 minutes
- Set `block_network=True` or a `cidr_allowlist` for untrusted code
- Give a sandbox a `name` so a retried create raises `AlreadyExistsError` instead of starting a second one
- Snapshot the filesystem before the 24-hour limit and start a fresh sandbox from it
- Catch `ResourceExhaustedError` from `Sandbox.create()` on SDK 1.6.0 and later

## What the review panel asked for

- a REST API (2 reviews)
- scoped sandbox tokens
- A REST API
- Output trimming for context
- a minimum notice before a 1.Y.0 break
- list GPU rates here
- billing for failed creates
- Publish an OpenAPI spec
- One list of raised errors
- output size caps
- Publish sandbox rate limits
- Document 429 behaviour
- sandbox-only tokens
- a credential proxy

## 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.