Massed Compute

# Fix list: Massed Compute

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/massed-compute, the October 2026 research run, assessed 9 October 2026. Grade E, 42.3 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 Massed Compute: 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, 20 out of 100, up to 16 more on the total

Why it scored 20: Graded on the hosted lines for the MCP server and the REST API. No status page was found on the site, the docs, the pricing page or the repository (0). No incident history to read (5). The troubleshooting page documents a 429 for request limits with no figures (0). The same page says to wait a minute and retry, and every tool carries an `idempotentHint`, but `instances_launch` has no idempotency key (5 of 15). The on-demand page says 99.999 per cent uptime is guaranteed. No SLA document was found, the End User Licence Agreement promises support only under an order form or SLA, and the Terms & Conditions could not be read (0). The API is `/api/v1`, the server card is version 1.0.0 and nothing is 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, 20 out of 100, up to 10 more on the total

Why it scored 20: No x402, MPP or L402 in the docs, the OpenAPI files, the server card or the pricing page. Crypto top-ups are made by the account owner in the console (0). Hourly prices for every on-demand configuration are public at vm.massedcompute.com/pricing without a login (20). No free tier or trial credit was found, and a first deposit is needed before a launch (0). Signup, email verification, billing setup and token creation all happen in a browser (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. Agent ergonomics, 56 out of 100, up to 7.2 more on the total

Why it scored 56: 17 tools whose definitions total about 10.5 KB in the server card, and a read-only credential cannot use the six write tools (15 plus 5 of 25). List tools take no arguments, with no pagination, filters or field selection. Only `recipes_search` has a `limit`, default 3 and maximum 10 (4 of 20). The docs explain 401, 402 and 429, the changelog names `-32003 Forbidden` for an out-of-scope call, and six deployment messages have written resolutions. None is in the OpenAPI file (10 of 20). All 17 tools carry `readOnlyHint`, `destructiveHint`, `idempotentHint` and `openWorldHint`. A launch has no idempotency key (14 of 20). A launch needs three fields and `regionName` accepts `any`. The only official packages are the npm and PyPI MCP wrappers, and the Go SDK and Terraform provider named on the product page were not found (8 of 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.

## 4. Security & auth, 61 out of 100, up to 6.8 more on the total

Why it scored 61: OAuth with discovery from a 401 challenge, or named API tokens shown once and revoked one at a time. Both come in two tiers, read-only and full access, with no finer scopes. Secrets travel only in a header (26 of 30). The read-only tier blocks launch, restart, terminate and SSH key changes. Destructive tool descriptions ask the model to confirm with the user, and a published skill adds a disclosure step. No server-side confirmation was found (13 of 20). The MCP server redacts VM passwords from tool results. Recipes are vendor-written Markdown that tools tell the model to follow step by step, and the docs give no guidance on untrusted content (9 of 15). Connected Apps lists OAuth grants. No audit log or per-call history was found (3 of 15). The wrapper's SECURITY.md gives a disclosure address, a two-business-day acknowledgement and a 30-day fix target, and its CI runs static analysis and dependency audits. The site claims SOC 2 Type II and HIPAA with no report or trust page, and there is no security.txt or bug bounty (10 of 20).

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.

## 5. Schema & documentation, 69 out of 100, up to 5 more on the total

Why it scored 69: A public OpenAPI 3.0 file covers the 14 REST operations, and a public server card lists all 17 MCP tools with JSON Schema inputs (25). massedcompute.com has an `llms.txt` that indexes marketing pages, and the vendor publishes an agent catalogue, the server card and four Markdown skills files. The docs sites have no `llms.txt` or Markdown copies (6 of 10). Tool descriptions state the purpose, the cost of a launch, when to search recipes first and when to confirm with the user. REST descriptions are one line each (16 of 20). Required fields are marked and extra properties are refused, but there are no enums, `productName` and `regionName` are free strings, and the OpenAPI file's required lists name `image_id` and `couponCode` where the properties are `imageId` and `coupon` (8 of 15). Every REST operation has a response example. Only 200 and 202 are documented, and errors appear only in a troubleshooting table and a deployment errors page (7 of 15). The path is versioned as `/api/v1` and the MCP wrapper has a dated changelog. No API changelog was found (7 of 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. Maintenance & community, 51 out of 100, up to 4.3 more on the total

Why it scored 51: The newest release of the MCP wrapper is v1.1.0 of 24 July 2026, 77 days before the check. The hosted server has changed since, with fields in the server card that the July definitions lack, but no dated API changelog records it (20 of 30). One release and one dated changelog entry in the last 90 days (0). A dated wrapper changelog and a Discord server named as the fastest support route. No API changelog, and the repository shows six open issues or pull requests whose answers we did not read (7 of 15). Listed in the official MCP registry as `io.github.Massed-Compute/mcp` at 1.1.0, published 31 July 2026, with current npm and PyPI packages (15). CI tests Node 20 and 22 and Python 3.10 to 3.13 on three operating systems, with Dependabot, static analysis and a nightly check of the hosted endpoint. The npm package has no runtime dependencies (9 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.

## 7. Transparency & trust, 61 out of 100, up to 3.4 more on the total

Made of editorial 53, provenance 68.

Why it scored 61: Closed service. The End User Licence Agreement of 15 September 2026 is clear and applies Nevada law, and the wrapper is MIT. The Terms & Conditions, the principal agreement, are drawn by a script we could not read (13 of 30). The privacy policy of 22 July 2026, the DPA of 27 March 2026 and the EULA are published. The EULA says VM data is destroyed at termination and persistent storage is kept at most 60 days after non-payment. The privacy policy is a general template that names a mobile application and gives no retention periods (19 of 30). No deprecation policy was found. The wrapper's changelog documents what 1.1.0 removed, and the older docs site still documents `api.massedcompute.com` while the current docs say not to use it (4 of 20). The sub-processor list names 23 processors with locations, is dated 10 September 2026 and has an RSS feed. The DPA says primary processing is in the United States (17 of 20).

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: massedcompute.com, registered 2021-10-28 (4 years) (7 of 15)
- Terms of service: published, but our reader couldn't read it (7 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)
- security.txt: not found (0 of 10)

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- 8 June 2026. The `instances_list` and `instances_get` MCP tools returned the cleartext VM password although their descriptions said it was redacted. Fixed in `massed-compute-mcp` 1.0.3 and later moved server-side, with rotation advice in the changelog. No GitHub advisory is published. Fixed and documented, so 3 points (https://github.com/Massed-Compute/massed-compute-mcp/blob/main/CHANGELOG.md).
- 9 October 2026. The product page lists SDKs for Go and a Terraform provider, webhooks and per-project quotas. None appears in the docs, the OpenAPI file or the vendor's public GitHub organisation. 2 points (https://massedcompute.com/products/inventory-api/).

## 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.

- The lead named api-docs.massedcompute.com as the docs. That site documents an older marketplace Inventory API on `api.massedcompute.com` with keys issued by support. The API an agent would use is `https://vm.massedcompute.com/api/v1`, documented at vm-docs.massedcompute.com. The MCP address in the lead was right.
- unchecked: the Terms & Conditions. The page embeds a Termly viewer drawn by script, so the principal agreement, any SLA, refund terms and any clause on automated access were not read. `provenance.terms` points at the page, and the policy reader may not be able to read it either.
- unchecked: any request to `https://vm.massedcompute.com/api/`. robots.txt on that host disallows `/api/`, so the live endpoint, its 401 challenge and its OAuth authorisation server metadata were not fetched. Tool definitions come from the public server card.
- unchecked: PyPI and npm download counts and the PyPI release dates. Release dates come from the repository's tags and changelog.
- unchecked: the replies on the repository's six open issues or pull requests.
- The tools reference says a read-only credential does not see write tools. The wrapper's changelog says the tool list is the same for every key and write calls are refused with `-32003`. We could not test which holds.
- The server card's authentication block mentions only Bearer API keys, while the docs describe OAuth as well.
- The sub-processor list names Amazon Web Services for infrastructure while the site says Massed Compute owns and runs all its hardware. Which workloads run on AWS is not stated.
- The SOC 2 Type II, HIPAA, GDPR and Tier III claims have no report, auditor or trust page linked.
- Whether the 8 June 2026 password fix reached npm and PyPI as 1.0.3 is unclear. The registry lists 1.0.2 and 1.1.0, and the repository tag for v1.0.3 is dated 10 July 2026.
- Recipe tool descriptions tell the model to call `recipes_search` and follow recipes step by step. Recorded as a fact about the tool set. No deduction taken.

## Weaknesses

- No status page, incident history or security.txt was found on any of the vendor's hosts
- A 429 answer exists but no rate limit figures are published, and the OpenAPI file documents no error response
- `POST /instance/launch` has no idempotency key, so a retried launch can start a second billed VM
- Until a fix dated 8 June 2026, `instances_list` and `instances_get` returned the cleartext VM password although their descriptions said it was redacted
- The product page lists Go and Terraform SDKs and webhooks that the docs, the OpenAPI file and the public GitHub organisation do not show
- A stopped instance is billed at the full hourly rate. Only termination ends billing, and it deletes the data

## 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.

- Use `https://vm.massedcompute.com/api/v1`. The older docs at api-docs.massedcompute.com describe a separate marketplace API on `api.massedcompute.com` whose keys come from support.
- Ask the owner for a read-only token unless the task must launch or terminate. Read-only credentials cannot call launch, restart, terminate or SSH key changes.
- Call `gpu_inventory_list` and `images_list` before `instances_launch`. Product names and image IDs are free values, and SXM-only images do not run on PCIe cards.
- Expect 402 on launch until the owner has added credit and set a recharge amount and threshold. List instances before retrying a launch.
- Terminate to end billing. Stopping keeps the full hourly charge, and termination deletes the instance data.

## 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.