# Fix list: Hume EVI (Empathic Voice Interface) From Anchor Terminal's listing at https://www.anchorterminal.com/tools/hume-evi, the October 2026 research run, assessed 1 October 2026. Grade C, 57.3 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 Hume EVI (Empathic Voice Interface): 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, 27 out of 100, up to 12.8 more on the total Why it scored 27: One API key and secret key pair per account, replaced as a whole with 'Regenerate keys', plus 30-minute access tokens from a client-credentials endpoint (15). The docs tell clients to put the API key or access token in the WebSocket URL as a query parameter, and the Twilio webhook URL carries the key the same way, so we deduct 10 (5 of 30). No scoped or read-only keys (3 of 20). EVI listens to callers and we found no prompt-injection guidance (5 of 15). Chat history keeps every session's events, and we found no audit log (8 of 15). The privacy page says Hume is HIPAA compliant and signs BAAs and DPAs on request. No security.txt, SOC 2 report or bug bounty found (6 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. ## 2. Reliability, 55 out of 100, up to 9 more on the total Why it scored 55: Statuspage at status.hume.ai with component history back to September 2025 (20). In the last 90 days, EVI was 'down in a majority of cases' and TTS impacted from 10:47 to 17:14 UTC on 11 July (about 6.5 hours, posted on 14 July), EVI error rates rose on 24 July for about 2.6 hours before chats recovered (resolved 27 July), and TTS was down for about 7 minutes on 19 September. Two majors sit between our one-major (10) and several-majors (0) lines, so we give 5. Limits are published, concurrent connections from 1 on Free to 30 on Business, 100 HTTP requests a second and a 30-minute session cap (15). The error catalogue has a rate-limit code (E0811, 'Retry later') and a too-many-chats code (E0700) that states the active count and limit, but no HTTP 429, Retry-After or backoff guidance (5 of 15). No SLA found (0). EVI 3 is the generally available default per the 30 September check (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. ## 3. Payments & pricing, 37 out of 100, up to 7.9 more on the total Why it scored 37: No x402, MPP or L402 (0 of 40). Plan prices and per-minute overage from $0.06 to $0.04 are on the public pricing page (20). The Free plan includes 5 EVI minutes a month for non-commercial use, and the billing docs say new accounts start on the free tier with $20 in credits. No page says whether a card is needed, but error E0300 ('Out of credits', activate billing to continue) suggests billing is a later step, so we give partial credit (17 of 20). Access starts with a human signup (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. ## 4. Agent ergonomics, 66 out of 100, up to 5.5 more on the total Why it scored 66: Scored as an API. Chat history and chat events are listed through paginated endpoints, and session messages are small (15 of 25). Paging on chats, chat groups and configs (16 of 20). Errors carry a code, slug and message with a documented recovery step, such as E0714 for an inactivity timeout and E0720 when a chat group can't be resumed because retention is off (18 of 20). No idempotency keys, and tool calls depend on a supplemental LLM (5 of 20). SDKs for Python, TypeScript and React, but EVI 3 has no default voice, so every config must name one (12 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. ## 5. Maintenance & community, 65 out of 100, up to 3.1 more on the total Why it scored 65: TypeScript SDK 0.16.1 and the EVI React SDK 0.3.0-beta.6 were tagged on 18 August 2026, 45 days ago (20). Three tagged releases since 4 July, 0.3.0-beta.5 on 8 July plus the two on 18 August, two of them beta tags of the React SDK (20). The public changelog stops at 15 May 2026 and we didn't test support (5 of 15 for a closed service). Python, TypeScript and React SDKs, the TypeScript SDK current and the Python SDK last tagged v0.14.0 on 17 June (12 of 15). MIT SDKs with ci.yml workflows in all three repos and Python 3.13 to 3.15 support (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. ## 6. Transparency & trust, 67 out of 100, up to 2.9 more on the total Made of editorial 48, provenance 86. Why it scored 67: Closed service with clear terms and MIT SDKs (18 of 30). The developer privacy page, re-read on 2 October, still says both that Hume 'will not use data submitted by customers via our API to train or improve our models' and that 'by default, anonymized data from user interactions with the EVI API is used to improve our models', with a toggle to opt out. Retention is on by default with a 'Do not retain data' toggle. The legal privacy policy (last updated 25 February 2025) says nothing about training and gives no retention periods (10 of 30). EVI 1 and EVI 2 reached end of support on 30 August 2025 with dated notices (15 of 20). Data is stored 'in the US and around the world' with no subprocessor list in either policy (5 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: hume.ai, registered 2020-04-06 (6 years) (11 of 15) - security.txt: not found (0 of 10) ## 7. Schema & documentation, 90 out of 100, up to 1.6 more on the total Why it scored 90: llms.txt links an AsyncAPI 2.6.0 file for the EVI WebSocket (28 message schemas, /chat and /chat/{chat_id}/connect channels, a typed Error message with code and slug) and OpenAPI files per API (25). llms.txt (10). The EVI docs explain configs, voices, supplemental LLMs and interruption settings with guidance on when to use each (15 of 20). Session settings and configs are typed, with enums and required lists in the AsyncAPI file and ranges such as `min_interruption_ms` from 50 to 2,000 (13 of 15). Examples in the SDK guides, and an errors page lists over 40 EVI and configuration codes and 17 TTS codes, each with a recommended action, plus WebSocket close codes 1000, 1008 and 1011 (15). EVI versions carry end-of-support dates, but the public changelog's newest entry is still 15 May 2026 (12 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. ## 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. - Which of the two training statements on the privacy page applies to EVI. The EVI-specific line says anonymised data trains models by default - Whether the free tier needs a card. No page says, and the billing docs ($20 in credits) and pricing page (5 EVI minutes) disagree on what it includes - The changelog announced EVI 4-mini on 3 October 2025, and we didn't recheck whether EVI 3 is still the default - We opened the EVI AsyncAPI file and the OpenAPI index, not the EVI OpenAPI file itself ## Weaknesses - One account-wide key, sent in the WebSocket and Twilio webhook URLs - Privacy page contradicts itself on training use of EVI data - EVI down in a majority of cases for about 6.5 hours on 11 July 2026, posted to the status page three days later - Public changelog ends at 15 May 2026 - No HTTP 429, Retry-After or backoff guidance, only 'retry later' error codes ## 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. - Create a config with a voice first, EVI 3 has no default - Pick a Claude, GPT, Gemini or Moonshot supplemental LLM if the agent needs tools, since Hume's own model can't call them - Mint 30-minute access tokens server-side so the API key never reaches a client URL - Turn on 'Do not retain data' and 'Do not use for training' at app.hume.ai before sensitive calls - Reconnect with the chat group ID before the 30-minute session cap, and on E0700 close an open chat before starting another ## What the review panel asked for - document 429 behaviour - post incidents as they happen - header auth on WebSocket and Twilio - 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.