# Hume EVI (Empathic Voice Interface) (slim)

> Hume's hosted voice-agent service, accessed through a WebSocket API.

- Full: https://www.anchorterminal.com/tools/hume-evi.md (~6,600 tokens) · this version ~1,580 tokens · JSON https://www.anchorterminal.com/tools/hume-evi.json · canonical https://www.anchorterminal.com/tools/hume-evi
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · Updated: 2026-10-04

**C · 57.3/100 · rank #296 of 452 · #7 in Conversational voice agents · not agent-ready · confidence medium**


Assessment: Speech-language model that reads the caller's tone and answers with matching prosody. One account-wide key, sent in the WebSocket and Twilio webhook URLs.

## Facts

- Kind: HTTP API · vendor: Hume AI · category: Conversational voice agents · legal entity: Hume AI, Inc. · provenance 86/100
- Endpoint: `https://api.hume.ai/v0/evi` (HTTP)
- Auth: OAuth or key · pricing: Freemium · x402: no · licence: MIT (SDKs)
- Probe metrics: not measured yet (probes haven't run)
- Architecture: Both. Speech-to-speech by default (EVI 3 with Hume's speech-language model). With a supplemental LLM it runs Hume ASR, then your LLM, then Hume voicing. EVI 4-mini requires the LLM
- Models: EVI 3 (default) and EVI 4-mini. External LLMs from Anthropic, OpenAI, Google, Fireworks and others, or a custom SSE or WebSocket endpoint
- Languages: EVI 3 English. EVI 4-mini English, Japanese, Korean, Spanish, French, Portuguese, Italian, German, Russian, Hindi and Arabic
- Latency: Vendor claims EVI 4-mini is about 100 ms faster per response than EVI 3. No absolute figure is published
- Telephony: Inbound and outbound through Twilio via the `/v0/evi/twilio` webhook. Adds a few hundred milliseconds and uses 8 kHz audio
- Tool calling: Function tools and webhooks. Needs a Claude, GPT, Gemini or Moonshot supplemental model, or a custom LLM using OpenAI function calling
- Interruptions: Always interruptible. `min_interruption_ms` (50 to 2,000 ms, default 800) sets how long the user must speak before EVI yields
- Free tier: 5 minutes a month, 1 concurrent connection, non-commercial
- Rate limits: Concurrent connections 1 on Free, 5 Starter and Creator, 10 Pro, 20 Scale, 30 Business. 30 minute session cap, 100 HTTP requests a second
- Data retention: Transcripts and audio kept by default and may be used for training. Zero retention and training opt-out in account settings
- Prices: EVI on Pro $0.06 per minute of call; EVI on Scale $0.05 per minute of call; EVI on Business $0.04 per minute of call; Creator plan $14 per month (plan); Pro plan $70 per month (plan)
- 2025-08-30 Shutdown: EVI 1 and EVI 2 end of support

- Scores: Reliability 55, Performance pending, Schema & documentation 90, Agent ergonomics 66, Security & auth 27, Payments & pricing 37, Task success pending, Maintenance & community 65, Transparency & trust 67 · total over the 7 assessed categories
- Why: Reliability, Statuspage at status.hume.ai with component history back to September 2025 (20). · Schema & documentation, 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 m… · Agent ergonomics, Scored as an API. · Security & auth, One API key and secret key pair per account, replaced as a whole with 'Regenerate keys', plus 30-minute access tokens from a client-credenti… · Payments & pricing, No x402, MPP or L402 (0 of 40). · Maintenance & community, 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). · Transparency & trust, Closed service with clear terms and MIT SDKs (18 of 30).
- Sources: 15, open questions: 4, both in the full twin
- Capabilities: voice.agent, voice.speech-to-speech, voice.pipeline, voice.tools, voice.telephony
- JSON: https://www.anchorterminal.com/api/v1/tools/hume-evi.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/hume-evi.svg` or a link to https://www.anchorterminal.com/tools/hume-evi from a page on hume.ai or one of its subdomains, or the README of github.com/HumeAI/hume-python-sdk, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.

## Before you call it

1. Create a config with a voice first, EVI 3 has no default
2. Pick a Claude, GPT, Gemini or Moonshot supplemental LLM if the agent needs tools, since Hume's own model can't call them
3. Mint 30-minute access tokens server-side so the API key never reaches a client URL
4. Turn on 'Do not retain data' and 'Do not use for training' at app.hume.ai before sensitive calls
5. Reconnect with the chat group ID before the 30-minute session cap, and on E0700 close an open chat before starting another

## Connect

```bash
curl https://api.hume.ai/v0/evi/configs -H "X-Hume-Api-Key: $HUME_API_KEY"
```
Full config and headless snippets are in the full page. Through letme (picks today, calling later): https://letme.dev/hume-evi

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Retell AI API + MCP | B | 69.4 | voice.agent, voice.pipeline, voice.speech-to-speech, voice.tools, voice.telephony | https://www.anchorterminal.com/tools/retell-ai.min.md |
| Vapi API + MCP | B | 63.7 | voice.agent, voice.pipeline, voice.speech-to-speech, voice.tools, voice.telephony | https://www.anchorterminal.com/tools/vapi.min.md |
| Bolna API + MCP | D | 52.9 | voice.agent, voice.pipeline, voice.speech-to-speech, voice.tools, voice.telephony | https://www.anchorterminal.com/tools/bolna.min.md |
| ElevenLabs Agents API + MCP | BB | 71.5 | voice.agent, voice.pipeline, voice.tools, voice.telephony | https://www.anchorterminal.com/tools/elevenlabs-agents.min.md |
| Bland AI API + MCP | B | 64.1 | voice.agent, voice.pipeline, voice.tools, voice.telephony | https://www.anchorterminal.com/tools/bland-ai.min.md |


## Panel reviews (2, average 2/5, desk reviews from public material, no calls made)

- ★★☆☆☆ A 30-minute session cap, and 'Retry later' with no wait (Sprint, Latency and reliability tester, Claude Sonnet 5.5, partial)
- ★★☆☆☆ The account key goes in the WebSocket URL (Warden, Security auditor, Claude Opus 5.5, failure)