# Fix list: Deepgram Voice Agent API From Anchor Terminal's listing at https://www.anchorterminal.com/tools/deepgram-voice-agent, the October 2026 research run, assessed 1 October 2026. Grade B, 68.4 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 Deepgram Voice Agent API: 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, 55 out of 100, up to 9 more on the total Why it scored 55: Statuspage at status.deepgram.com with per-component incidents and a feed back to 12 May 2026 (20). Fifteen incidents since 3 July 2026, several of them an hour or more on parts the agent socket depends on. Flux STT streaming errors for about 2.5 hours on 7 July, failed Voice Agent responses on unpinned Gemini models for about 1.5 hours on 21 July, batch and streaming STT degraded for about 3.5 hours on 4 August, and Flux TTS errors on the global endpoint for about 4 hours on 25 September (0). Deepgram writes up short incidents most vendors wouldn't post, which counts against it here and in its favour under transparency. Concurrency is published, 45 sockets on pay as you go, 60 on Growth in North America and from 100 on Enterprise (15). Over-limit requests get a 429 and the docs recommend exponential backoff, with no Retry-After header and no idempotency guidance for function calls (10 of 15). The pricing page lists 'Standard Uptime' on self-serve plans with no figure, and we found no SLA terms (0). The Voice Agent API 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. ## 2. Payments & pricing, 40 out of 100, up to 7.5 more on the total Why it scored 40: No x402, MPP or L402 (0 of 40). Per-minute prices for every tier on the public pricing page (20). $200 of credit with no card (20). The first key has to come from a human signup in the Console, though later keys can be made through the API (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, 67 out of 100, up to 5.4 more on the total Why it scored 67: Scored as an API. Events on the socket are small and typed, function-call context can be trimmed, and the latency report and history are opt-in, but there's no field selection on the REST side (18 of 25). Pagination and filtering matter little on a socket, and the REST endpoints for models and saved agent configurations are small lists (10 of 20). Errors and warnings are documented as typed events (14 of 20). The function-call hold added on 8 September 2026 defers irreversible functions until the user's turn is confirmed, but there are no idempotency keys (10 of 20). Sensible defaults (Nova-3 if no STT is named) and official SDKs in Python, JavaScript, Java, Go, C# and React (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, 71 out of 100, up to 5.1 more on the total Why it scored 71: API keys carry an owner, admin or member role, can be set to expire, can be created through the API and can be tagged, and browsers get 30-second JWTs from the token endpoint (30). Member keys limit what a leaked key can do, and the function-call hold keeps the agent from firing irreversible tools on a speculative turn, but there's no read-only agent mode (14 of 20). The agent feeds caller speech to an LLM and we found no prompt-injection guidance (5 of 15). Requests are tagged per key and each session reports latency and history, but we found no audit log of account actions (10 of 15). The vendor states SOC 2 Type 1 and Type 2, HIPAA and PCI DSS. No security.txt and no bug bounty found (12 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. Transparency & trust, 80 out of 100, up to 1.8 more on the total Made of editorial 70, provenance 90. Why it scored 80: The service is closed with clear terms, and the official SDKs are MIT (18 of 30). The data page says audio and transcripts are kept for model improvement unless `mip_opt_out` is set, in which case they're deleted after processing, and it states that a full in-region guarantee needs both a regional endpoint and the opt-out. The statements agree, though keeping data is the default (22 of 30). The changelog dates changes and corrections, but we didn't find a published deprecation policy for the agent API (10 of 20). A subprocessor page and EU, India and Australia endpoints are published (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): - security.txt: not found (0 of 10) ## 6. Schema & documentation, 90 out of 100, up to 1.6 more on the total Why it scored 90: Public OpenAPI and an AsyncAPI file that describes the agent socket messages (25). llms.txt with a Markdown copy of each page, and a keyless docs MCP server (10). Every client message and server event has its own page saying what it's for, from `Settings` to `ForceEndTurn` and `FunctionCallRequest` (17 of 20). The AsyncAPI types the Settings message with enums and required fields, though provider `endpoint.headers` are free-form (13 of 15). Examples in every language guide and a Voice Agent errors and warnings page, which we read only in outline (11 of 15). A dated changelog that also publishes corrections, such as the token TTL field name on 18 September 2026 (14 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. ## 7. Maintenance & community, 85 out of 100, up to 1.3 more on the total Why it scored 85: Changelog entry on 30 September 2026 (30). Over a dozen dated entries since July, several on the Voice Agent API (20). A public changelog that issues corrections, plus support channels we didn't test (12 of 15 for a closed service). SDK releases on 14 September 2026 (Python 7.9.0, JavaScript 5.11.0, Java 0.10.0) added Voice Agent support for `ForceEndTurn` (15). MIT SDKs with current releases (8 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. - We read the Voice Agent errors and warnings page only in outline. - Whether opting out of the Model Improvement Program changes the price. - We didn't re-check security.txt on deepgram.com and kept last week's finding of none. ## Weaknesses - No phone numbers or SIP, so you bridge Twilio or another carrier yourself - Fifteen status incidents since July, four of them an hour or longer on parts the agent uses - Call audio is kept for model improvement unless `mip_opt_out` is set - No SLA terms on self-serve plans - Sessions end at 2 hours ## 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 `mip_opt_out` in the Settings message to keep call audio out of training - Use the function-call hold for anything irreversible, since replies can start before the user's turn is confirmed - Back off exponentially on 429, a pay-as-you-go project gets 45 concurrent sockets - Pin the LLM version, an unpinned Gemini route failed for 1.5 hours on 21 July 2026 - Mint browser tokens with `ttl_seconds`, not `ttl`, or they expire after 30 seconds ## What the review panel asked for - publish SLA terms - add Retry-After to 429 - opt-out as the default ## 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.