Replicate Deployments by Replicate

# Fix list: Replicate Deployments

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/replicate-deploy, the October 2026 research run, assessed 1 October 2026. Grade B, 63.7 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 Replicate Deployments: 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. Security & auth, 40 out of 100, up to 10.5 more on the total

Why it scored 40: Bearer tokens starting `r8_`, several per account, each can be disabled; no scopes or expiry (20). No read-only or per-model token; every token can create, update and delete deployments (0). Returns your own model's output (10). No audit log or per-token usage view found; predictions are listed per account (5). GitHub secret scanning disables leaked tokens and emails the owner; no security.txt, bug bounty or certification found in the docs we read (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.

## 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 prices for each hardware SKU published without a login (20). Accounts without a card can run predictions at up to 6 a minute; private deployments still bill set-up and idle time (10 of 20). Signup is a browser flow (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, 68 out of 100, up to 5.2 more on the total

Why it scored 68: The MCP server exposes one tool per HTTP operation and has an experimental code mode that collapses them into two tools (search the SDK docs, run TypeScript) (20). List pagination not verified this run; no field selection (10). Coded errors with suggested fixes, `detail` messages on HTTP errors (15). No idempotency keys; `Prefer: wait`, webhooks and cancel cut polling, and a failed run still bills its active time (8). Sensible defaults (`min_instances` 0 allowed) and official Python and JavaScript clients plus Cog (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. Reliability, 75 out of 100, up to 5 more on the total

Why it scored 75: Replicate incidents now post under a Replicate component on cloudflarestatus.com, with history (20). Four incidents in September 2026, all marked minor by Cloudflare. Some third-party models couldn't scale out for 15 hours 41 minutes on 14 and 15 September, a Pruna-specific issue ran 20 hours on 17 September, backend services returned intermittent 500s for 1 hour 54 minutes on 24 September, and hot-swapped Flux models stuck for 17 minutes on 28 September; under our rule that's minor incidents only (20). 600 prediction creates a minute, 3,000 a minute on other endpoints, and 6 a minute without a card (15). A 429 body says when the limit resets ('resets in ~30s') and the error-code page gives retry advice per code, but there's no Retry-After header or idempotency guidance (10 of 15). No SLA found (0). Deployments are GA (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.

## 5. Maintenance & community, 70 out of 100, up to 2.6 more on the total

Why it scored 70: Cog 0.23.0 released on 22 September 2026 per last week's research (30). Release cadence for Cog not re-checked this run, and the changelog has no entries since April (10). We didn't review Cog issue response times this run (10 of 25). The changelog records auto-discovery through the MCP Registry from 10 February 2026, and Python and JavaScript clients are current (15). Package health not audited (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.

## 6. Schema & documentation, 85 out of 100, up to 2.4 more on the total

Why it scored 85: Public OpenAPI at api.replicate.com/openapi.json covering deployments, hardware, models and predictions (25). llms.txt with Markdown pages (10). Operation descriptions in the spec explain purpose with curl examples; deployment docs say when to use a deployment rather than a model version (16). Deployment fields are bounded (`min_instances` 0 to 5, `max_instances` 0 to 20, a 64-character version, a hardware SKU from GET /v1/hardware) (13). Examples throughout and 8 coded errors (E1001 out of memory, E6716 start timeout and others) with fixes; the HTTP error body format isn't described (11). Versioned /v1, but the public changelog's last entry is 21 April 2026 (10).

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.

## 7. Transparency & trust, 80 out of 100, up to 1.8 more on the total

Made of editorial 69, provenance 90.

Why it scored 80: Closed service; Cog is Apache-2.0 (20). API prediction inputs, outputs, files and logs are deleted after one hour by default, web predictions are kept until deleted, and a subprocessor page is published (22). Dated deprecations in the changelog (streaming default in July 2024, spend limits in July 2025) but no written policy (12). Subprocessors listed; data locations not stated in what we read (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):

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

## 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.

- replicatestatus.com served an old page last updated in April when we fetched it; we couldn't confirm the redirect to Cloudflare's status page the previous listing described.
- We couldn't re-check Cog's release history or issue tracker this run because of fetch limits.
- No certification (SOC 2 or similar) or disclosure policy was found in the docs we read; Cloudflare's programmes may now cover Replicate, but we found nothing saying so.

## Weaknesses

- Private instances bill set-up and idle time, H100 at $5.49 an hour
- API tokens have no scopes, expiry or audit log
- Changelog silent since 21 April 2026
- Only T4, L40S, A100 and H100, and more than 2 GPUs needs a committed-spend contract
- Two September 2026 incidents ran 15 and 20 hours, both marked minor

## 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.

- List `GET /v1/hardware` first and use the returned `sku` in the deployment body
- Set `min_instances` to 0 for bursty work; a warm H100 bills $5.49 an hour whether called or not
- Send `Prefer: wait` on deployment predictions to block instead of polling
- Copy outputs within an hour; API prediction data is deleted after that
- Wait for the reset time in the 429 body before retrying; prediction creates cap at 600 a minute

## What the review panel asked for

- publish the idle window length
- put multi-GPU prices on the page
- Send a Retry-After header
- Publish an SLA

## 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.