# Fix list: Agent 37 Cloud

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/agent37, the October 2026 research run, assessed 8 October 2026. Grade D, 46.1 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 Agent 37 Cloud: 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, 27 out of 100, up to 14.6 more on the total

Why it scored 27: Graded as a hosted service. No status page found, status.agent37.com returns 404 and neither the site nor the docs link one (0). No readable incident history (5). No request rate limits are published. The only numbers found are one on-demand backup per instance every 15 minutes and three concurrent template builds (0). Retry guidance is written per code, with backoff for `try_again`, `no_capacity` and 429, a Retry-After header on the backup limit, an idempotency key on budget top-ups and a warning that a timed-out exec keeps running, but instance create and exec take no idempotency key (12 of 15). The terms promise no uptime level unless agreed in writing (0). The Cloud API carries no beta label, though it launched on 29 September 2026 (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, 20 out of 100, up to 10 more on the total

Why it scored 20: No x402, MPP or L402 on the API (0). Per-unit prices for vCPU, memory and disk are published without login (20). The one-time $5 credit is granted only after a card is added, so there is no free tier without a card (0). A person signs up, adds the card and creates the key in the dashboard, and the docs say there are no balance or billing endpoints (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.

## 3. Security & auth, 49 out of 100, up to 8.9 more on the total

Why it scored 49: One workspace-wide `sk_live_` key, revocable, stored as a one-way hash per the privacy policy, with an optional IP allowlist of up to 20 addresses or ranges. No scopes were found (20). Signed URLs carry an expiring token in the query string for browser access. We don't deduct for that, as with presigned URLs elsewhere, but the token can't be revoked and can last seven days. No read-only keys and no confirmation on delete or restore. Exec runs as the image user unless root is requested, each instance has a managed-spend cap, and an agent has no route to open a public port (8). The agent returns untrusted content. The docs warn about serving agent-written files inline and about empty channel allowlists, and the vendor says sandboxes run under gVisor with egress limited to the public internet (10). A usage endpoint breaks spend down by instance and model, and container logs are readable, but no audit log of API calls was found (5). The vendor says a SOC 2 Type I report is available under NDA. The trust centre refused our reader, and no security.txt, disclosure policy or bug bounty was found (6).

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, 57 out of 100, up to 7 more on the total

Why it scored 57: No OpenAPI or other machine-readable contract found. The paths we tried return 404 and the 637 KB llms-full.txt never mentions one (0). llms.txt with 49 pages, llms-full.txt and a Markdown copy of every page (10). Pages say what each call is for and when to choose another, such as stop against delete, resize against fork and exec for anything unwrapped (17). Parameters are documented with types, ranges, defaults and patterns in prose, with no schema to validate against (10). Curl, Python and JavaScript examples with responses on each page, and an error reference listing 18 Hosting API codes, 13 transport codes and 16 Agent API codes (15). The path is versioned at /v1, template images carry immutable dated tags and the gateway reports a semver, but no public changelog was found (5).

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. Agent ergonomics, 60 out of 100, up to 6.5 more on the total

Why it scored 60: Exec output is capped at 512 KB a stream with a `truncated` flag, logs take a `tail` and chat can return one JSON body instead of a stream, but there is no field selection and `GET /v1/instances` returns every full object (12). No pagination or filters were found on the instance, session or file lists. Usage and metrics take date or hour windows (8). Errors carry stable codes, with `param`, `hint` and `response_id` on the Agent API and a remedy per code (20). An idempotency key on budget top-ups, `X-Expected-Mtime` and `overwrite=false` on file writes and documented safe retries, with none on create or exec (12). An empty create body works and every field is optional, but there is no SDK library, only a CLI and raw HTTP examples (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.

## 6. Transparency & trust, 56 out of 100, up to 3.9 more on the total

Made of editorial 56, provenance 56.

Why it scored 56: Closed service with dated terms (4 August 2026). Template Dockerfiles and example apps are MIT and the CLI is Apache-2.0 (15). The privacy policy (1 September 2026) gives retention by category, seven days for a deleted instance's recovery copy, up to 30 days for request logs and about seven years for billing records, and the docs repeat the seven-day window. No DPA was found (22). One deprecation is documented, the Bearer header on instance URLs, with an email promised before removal and no date. No written policy (5). The policy names 12 kinds of third-party service and the docs say instances run on dedicated servers in the United States. The hosting providers aren't named (14).

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: agent37.com, registered 2025-04-29 (1 year) (3 of 15)
- Terms of service: read, states 6 of the 7 things a reader expects (9.1 of 10)
- Privacy policy: read, states 7 of the 8 things a reader expects (9.3 of 10)
- Status page: not found (0 of 10)
- Changelog: not found (0 of 10)
- security.txt: not found (0 of 10)

## 7. Maintenance & community, 65 out of 100, up to 3.1 more on the total

Why it scored 65: Scored on the closed-service scale. The newest template images are tagged 2026.10.05a, three days before the check (30). Hermes image tags on 27 September and 2, 3 and 5 October 2026, and CLI releases 0.3.0, 0.3.1 and 0.4.0 between 3 August and 11 September (20). No public changelog. Support is by email, a dashboard chat and a founder's booking link, and we couldn't read GitHub issues (5 of 15). No SDK library. The `agent37` CLI on npm is current at 0.4.0 (5 of 15). The templates repository has a test workflow, whose results we couldn't read (5 of 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.

- unchecked: the trust centre at trust.agent37.com answered 403, so the SOC 2 Type I report, subprocessor list and any disclosure policy there are unread
- unchecked: GitHub star counts and open issues for agent37-platform/templates and examples. The GitHub API rate-limited our address, so `githubStars` is null
- unchecked: no account was opened and no instance created, so isolation, wake times, key revocation in the dashboard and whether keys can be limited beyond the IP allowlist are from the docs only
- Whether request rate limits exist. None are published
- Whether the first release date of the Cloud API is earlier than the npm CLI's first version of 16 July 2026. `firstReleased` is left null
- The legal entity is written Agent 37 Inc. in the terms and Agent37.com Inc in the privacy policy's representative section
- The scout's entry held up on every fact we rechecked. The launch post's $2,500 of credits is an offer to YC readers and isn't on the pricing page, where the public credit is $5

## Weaknesses

- No status page was found. status.agent37.com returns 404, and the terms promise no uptime level
- No OpenAPI file, SDK library or public changelog was found. The npm package `agent37` is a CLI
- No API rate limits are published beyond one on-demand backup per instance every 15 minutes
- One workspace-wide key controls every instance. No scopes or read-only keys were found
- The $5 starter credit needs a card on file, and keys and top-ups exist only in the dashboard
- The launch is dated 29 September 2026 and the public templates repository starts on 2 October 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.

- Send the key as `Authorization: Bearer` to api.agent37.com and as `X-Agent37-Key` to the instance URL. The Bearer form on instance URLs is deprecated
- Poll `GET /v1/health` on the instance URL for `"healthy": true` before the first message. `status: running` only means the machine is up
- Check whether `error` is a string before reading `error.code`. Transport errors are flat strings, and a failed turn returns 200 with `status: "failed"`
- Don't blindly retry an exec that timed out after 13 minutes. The command keeps running, so a retry starts a second copy
- Set `auto_sleep: true`, keep anything that must survive a sleep in a file, and allow a first-byte timeout of three minutes for wakes

## 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.