Cerebrium

# Fix list: Cerebrium

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/cerebrium, the October 2026 research run, assessed 8 October 2026. Grade C, 55.3 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 Cerebrium: 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, 48 out of 100, up to 10.4 more on the total

Why it scored 48: Graded as a hosted service. Status page at status.cerebrium.ai with 11 components and 90 days of history (20). Since 10 July 2026 the Inference API had a 10-minute outage on 21 July and degradations of 2 hours 19 minutes on 15 July, 3 hours 13 minutes on 16 July, 5 hours 7 minutes on 19 August (US, upstream) and 7 hours 20 minutes on 2 September (EU, provider network), which we read as one major outage's worth and not a clean record (10). No request rate limits found. Plans publish concurrency caps of 5 GPUs and 500 containers on Hobby and 30 and 1,000 on Standard (5 of 15). No 429 or backoff guidance for callers found. Webhook forwarding retries failed deliveries and the docs ask for idempotent webhook receivers (3 of 15). No SLA found. The 99.9 and 99.999 per cent uptime figures on the site are claims without terms (0). Endpoints, the CLI and the REST API are GA, while multi-region deployment and checkpointing are marked 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, 30 out of 100, up to 8.8 more on the total

Why it scored 30: No machine payment protocol (0). Per-second GPU, CPU and memory prices and the storage price are published without a login (20). The Hobby plan has no fee, but compute is billed from the first second, no free compute allowance is stated and the pricing page does not say whether a card is needed (10 of 20). Signup and `cerebrium login` are browser flows (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, 49 out of 100, up to 8.3 more on the total

Why it scored 49: No field selection. A `limit` parameter on five list operations, and app and run objects are small (12 of 25). The audit log and runs page with `limit` and `nextToken` and filter by status, status code and time, though other lists use `page`, `pageSize` or `offset` (14 of 20). Errors are a `message` string with 400 or 401. Deployed endpoints return 200, 401, 403, 404 or 5xx, and a function can set its own `status_code` (10 of 20). No idempotency keys. Async calls return a `run_id` that can be listed and cancelled (5 of 20). Defaults are sensible (`min_replicas` 0, `cooldown` 60) and `cerebrium init` writes a working config, but there is a CLI only and no language SDK (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, 60 out of 100, up to 7 more on the total

Why it scored 60: Service account tokens expire after up to one year (or never, if set to 0), are limited to 1 to 50 named projects and are revoked by deleting the account. Inference API keys are separate and can be deleted. No per-action scopes, and the CLI's SECURITY.md says tokens sit in plaintext in a config file with mode 0644 (20 of 30). Users have owner, member and billing roles, but tokens have no read-only form, and `disable_auth` defaults to true, so deployed endpoints answer without a token unless the owner changes it (5 of 20). Runs the owner's code and returns its output (10). Audit log of 22 actions with actor, IP address and outcome, in the dashboard and over the API, on Standard (7 days) and Enterprise (30 days) only (12 of 15). SOC 2 Type II announced on 8 July 2026, HIPAA BAA on request, a trust centre, and disclosure to security@cerebrium.ai. No security.txt and no bug bounty found (13 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, 70 out of 100, up to 4.9 more on the total

Why it scored 70: Public OpenAPI 3.0 spec for the management API with 94 operations on 73 paths. Deployed endpoints take whatever the owner's function defines (25). llms.txt and a .md twin of every docs page (10). Operation descriptions are one line each. The guides say when to use the protected tier, async runs or a custom runtime (12 of 20). Required fields are marked, but the spec has no enums (valid values such as member and billing sit in prose) and several arrays and objects are untyped (7 of 15). The guides carry many worked examples. The spec has no examples and documents only 400 and 401 errors with a `message` string (8 of 15). Paths are versioned v2, v3 and v4 and CLI releases are dated on GitHub, but the spec version is a fixed 1.0.0 and no platform changelog was found (8 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. Transparency & trust, 63 out of 100, up to 3.2 more on the total

Made of editorial 45, provenance 80.

Why it scored 63: Closed service with published terms. The CLI is MIT (20). The docs say request and log data is deleted after 7 days on Hobby and 30 on Standard, that customer data is not used for training, and that a DPA is available on request. The privacy policy gives no retention periods, the pricing page shows both 1 day and 7 days of log retention for Hobby, and the docs still say SOC 2 Type I (15 of 30). No deprecation policy or dated notices found (0). Regions and some providers (AWS regions, Crusoe, Nebius) are named in the docs and status page, and Datadog, BugSnag and Stripe are named in passing, but no subprocessor list was found. The CLI's crash reporting is on by default with a documented opt-out (10 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: cerebrium.ai, registered 2021-06-11 (5 years) (11 of 15)
- Terms of service: read, states 5 of the 7 things a reader expects, and has 2 clauses that cost points (4.3 of 10)
- Privacy policy: read, states 7 of the 8 things a reader expects (9.3 of 10)
- security.txt: not found (0 of 10)

## 7. Maintenance & community, 75 out of 100, up to 2.2 more on the total

Why it scored 75: CLI 2.9.0 on 16 September 2026, 22 days before the check (30). Five releases since 7 August 2026 (2.5.4, 2.6.0, 2.7.0, 2.8.0, 2.9.0) (20). Closed service with a Discord community and Slack or Intercom support by plan. No platform changelog, and we couldn't read the GitHub issue tracker (8 of 25). The CLI is current on PyPI, Homebrew and as binaries. There is no language SDK (10 of 15). The repository has lint and test workflows and pins Go 1.25.2. CI status wasn't visible in the clone (7 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: GitHub stars and the issue tracker for CerebriumAI/cerebrium. The GitHub API refused us for its rate limit, so `githubStars` is empty and issue responsiveness is scored low.
- unchecked: trust.cerebrium.ai is drawn by script, so the compliance reports and any subprocessor list there were not read.
- Whether a new Hobby account needs a card or receives free compute credit. The pricing page and docs don't say, and the dashboard signup wasn't tried.
- The pricing page shows both 1 day and 7 days of log retention for Hobby, and the docs security page says SOC 2 Type I while the 8 July 2026 post says Type II.
- The terms of service describe themselves as governing the web pages at www.cerebrium.ai, yet they are the only terms published and cover accounts and subscriptions. Neither they nor the privacy policy carry a date.
- First release date of the CLI. The oldest version on PyPI is 1.18.10 from 27 March 2024, which is not the first.

## Weaknesses

- `disable_auth` defaults to true, so a deployed endpoint answers without a token unless the owner changes it
- No request rate limits, 429 guidance, idempotency keys or SLA found in the reviewed documentation
- Several multi-hour degradations of the Inference API between 15 July and 2 September 2026, and a 10-minute outage on 21 July
- No deprecation policy, platform changelog or public subprocessor list found
- The CLI stores tokens in plaintext in `~/.cerebrium/config.yaml` with mode 0644, per its own SECURITY.md

## 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.

- Set `disable_auth = false` in `cerebrium.toml` before deploying. The default leaves the endpoint callable by anyone with the URL
- Authenticate headless with `CEREBRIUM_SERVICE_ACCOUNT_TOKEN`. `cerebrium login` opens a browser
- Raise `response_grace_period` for long work. It defaults to 15 minutes and async runs stop at 12 hours
- Send `?async=true` to get a `run_id` with HTTP 202, and add `webhookEndpoint` because async calls return no result to the caller
- Check the plan before choosing hardware. A100, H100, H200, B200 and RTX PRO 6000 need Standard, and `protected` compute bills at twice the listed rate

## 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.