Cartesia Voice Cloning API + MCP by Cartesia

# Fix list: Cartesia Voice Cloning API + MCP

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/cartesia-voice-cloning, the October 2026 research run, assessed 1 October 2026. Grade C, 59.8 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 Cartesia Voice Cloning API + MCP: 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, 10 out of 100, up to 11.3 more on the total

Why it scored 10: No x402, MPP or L402 (0). Plan prices are public and cloning is included from Pro at $5 a month, but there's no dollar price per unit, only credits (10). The Free plan has no cloning (0). Signup is a human 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.

## 2. Security & auth, 48 out of 100, up to 9.1 more on the total

Why it scored 48: Graded for voice cloning, with consent and misuse controls in place of the read-only line and training and retention of voice data in place of the prompt-injection line, since the API returns audio and IDs rather than third-party text. Revocable API keys with a separate admin key, and short-lived access tokens with TTS, STT and agent grants for browser use, but no key scoped to cloning (25 of 30). No consent or speaker verification in the clone API, the Terms ask for the speaker's express permission, and no watermark or detection tool found (0 of 20). The Terms allow training on inputs by default, with an opt-out by online form, and no retention period for samples beyond a post-termination window in the DPA (5 of 15). Usage totals in the dashboard and through an admin-key endpoint, no per-call log found (8 of 15). No security.txt, the path returns 404, and no disclosure policy or bug bounty found, SOC 2 Type II, HIPAA and PCI claimed in Cartesia's own GDPR post, and a trust centre at trust.cartesia.ai (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.

## 3. Reliability, 63 out of 100, up to 7.4 more on the total

Why it scored 63: Status page at status.cartesia.ai (incident.io) with a Voice Cloning component and history (20). The last 90 days hold one major incident, elevated errors across the whole API including Voice Cloning for 58 minutes on 1 August, plus a run of smaller ones, among them a 41 minute partial TTS outage in the US on 29 July and about 21 hours of degraded cloning in APAC on 3 and 4 September (10). Concurrency limits per plan are published, 2 to 15 TTS requests (15). A 429 is documented when the limit is hit, with no Retry-After or backoff guidance in the docs, but the official SDK README says it retries 429 and 5xx twice with exponential backoff (8). No idempotency key or safe-retry guidance for clone creation (0). No SLA published (0). Instant cloning is generally available (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.

## 4. Agent ergonomics, 65 out of 100, up to 5.7 more on the total

Why it scored 65: Responses are compact voice objects, but we found no filter to list only your own voices, and the MCP server loads 19 tools (20 of 25). Voice lists paginate, with auto-paginating SDK iterators, filters not confirmed (10 of 20). Errors are HTTP statuses mapped to SDK exceptions, without cloning-specific codes (10 of 20). No idempotency key. Pro clone fine-tunes expose a status to poll (10 of 20). Three required fields plus the version header, and official Python and TypeScript SDKs (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.

## 5. Transparency & trust, 71 out of 100, up to 2.5 more on the total

Made of editorial 60, provenance 82.

Why it scored 71: Closed service with clear terms, SDKs under Apache-2.0 (15 of 30). Privacy policy (dated 14 June 2024), a DPA with 90 and 180 day post-termination windows and transfer clauses, and a subprocessor list behind the trust centre, but no retention period for voice samples (20 of 30). Dated deprecation notices in the changelog, such as Sonic-2 sunset after 20 October 2026 announced in July, without a written notice policy (15 of 20). Subprocessors sit in the trust centre, which we couldn't read without a browser, and regional endpoints are documented for Enterprise, while the privacy policy names no data locations (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: cartesia.ai, registered 2023-05-10 (3 years) (7 of 15)
- security.txt: not found (0 of 10)

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

Why it scored 85: OpenAPI files for the latest and each dated API version are listed in llms.txt (25). llms.txt and Markdown pages served (10). The cloning guides say when to use an instant clone and when to move to a Pro clone (14 of 20). Clone inputs are typed, with a 51-code language enum, an access enum, a 32 character tagline limit and accepted file formats (13 of 15). Examples in curl and both SDKs, but no error responses documented on the clone endpoint, only SDK exception classes (8 of 15). Dated `Cartesia-Version` header and a monthly changelog (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.

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

Why it scored 80: cartesia-mcp 0.26.0 on 2026-09-30 and SDK 4.2.0 on 2026-09-02 (30). More than three releases in the last 90 days, SDK releases on 29 July, 14 August, 26 August and 2 September (20). Public changelog, but we didn't confirm a support channel that answers (5 of 15). Current official Python and TypeScript SDKs (15). The Python SDK supports 3.9 to 3.14 and shipped within 30 days (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.

- The OpenAPI files are listed in llms.txt by path only, we didn't confirm the full URL, so the listing's `openapi` stays null
- The trust centre and its subprocessor list need a browser, we couldn't read them
- Whether the voice list can be filtered to the caller's own clones
- Whether the Playground asks for a consent attestation when cloning, the API doesn't

## Weaknesses

- No consent or speaker verification in the clone API
- Terms allow training on uploads by default, with opt-out only by form
- Zero Data Retention doesn't cover voice cloning
- No cloning on the Free plan and no per-unit dollar price
- No security.txt

## 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 `Cartesia-Version` on every call. The clone endpoint lists it as a required header
- Pass the speaker's recorded `language`. A clone is built from one language, add others with the accents API
- Keep IVC uploads under 16 MB
- Don't blind-retry `POST /voices/clone`. The SDK retries 429 and 5xx twice and there's no idempotency key, so check the voice list first
- Poll a Pro clone fine-tune until it finishes before using the voice ID

## What the review panel asked for

- Idempotency key on clone
- Own-voices filter
- a consent record on clone requests
- cloning-scoped keys

## 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.