One API to understand how people feel — in calls, messages and AI conversations. Analyse emotion, build long-term emotional profiles and catch cloned voices, then plug it into your apps, CRM, BI or AI agents.
Built for everyone from solo developers and SMEs to the world's largest platforms.
// Authorization: Bearer snt_live_•••• { "summary": { "sentiment": "mixed", "emotional_delta": { "delta": 0.70 }, "engagement": { "engagement_score": 71 }, "authenticity": { "synthetic_voice_risk": 0.04 }, "eri": { "score": 27.4, "band": "elevated", "vectors": { "frustration": 0.41, … } } }, "customer": { "emotions": [ { "label": "concern", "source": "acoustic", "share": 0.21 } ] }, "agent": { "aehs": { "score": 58.3 } } }
Use one capability or all five. Every response is structured JSON — ready for your database, your dashboards or your AI agent.
Calls, voice notes, chat, email, reviews and AI conversations — emotions, behaviours, engagement and change.
Explore →A personal emotional baseline for every customer, user or player — and every rep or bot.
Explore →Deepfake and cloned-voice risk scoring, plus the Kerr Test for live calls.
Explore →Aggregates and exports for Power BI, Tableau, Looker, Snowflake and your own models.
Explore →Tool definitions and an MCP server give any LLM agent emotional awareness.
Explore →POST /analyses/audio or /analyses/text. Pre-signed URL, direct upload or JSON — authenticated with your API key.
A signed analysis.completed webhook, or poll. ERI, F/A/C vectors, emotions, behaviours and AEHS as structured JSON.
GET /profiles/customers/{ref} for the emotional fingerprint. GET /insights/aggregates for your dashboards.
If people talk, type or speak to your AI, Sentient helps you understand them — in any industry.
Coach every rep on every call; spot hesitation and objections.
Emotional warmth on every CRM contact and account.
Emotion by journey step, product and channel.
QA at 100% coverage and targeted coaching.
Adapt tone and hand off at the right moment.
Emotion and profiles as a tool call or MCP server.
Spot cloned and AI-generated voices across 16 high-risk scenarios. See below ↓
Pressure and urgency patterns routed to fraud teams.
Emotion timelines across interviews and tests.
Reviews, surveys and social, scored and trended.
Feedback on tone; mood check-ins (non-diagnostic).
Player frustration and toxicity signals.
Voice cloning takes seconds and fools people and voiceprints alike. Sentient scores calls, voice notes and audio clips for synthetic-voice risk — and checks what clones still get wrong: genuine emotional reaction.
Cloned voices calling to change details, reset access or move money.
“Urgent transfer” calls and voice notes impersonating executives.
An extra authenticity layer on top of voiceprint log-in.
Checks before withdrawals, recoveries and limit changes.
Attackers posing as staff to reset passwords or MFA.
Screen the audio of Teams, Zoom and Meet calls for cloned voices.
Synthetic voices used by fake candidates or new starters.
“Mum, I need help” voice notes in messaging apps.
AI voices used to build false trust on dating and selling platforms.
Voice checks before number ports, SIM swaps and plan changes.
Claims phoned in with a borrowed or cloned voice.
Check viral audio clips before they’re published or shared.
Flag synthetic audio of public figures at scale.
Detect synthetic audio uploads on social and UGC platforms.
Protect tax, benefits and healthcare lines from impersonation.
Risk indicators on audio submitted in disputes and investigations.
Ship an emotion feature this afternoon.
Insight without a data team.
The emotion engine inside your product.
Voice clones can fool people and voiceprints. Sentient checks what fakes still get wrong — how a real person emotionally reacts — alongside the acoustic traces of synthetic speech. For high-risk moments, the Kerr Test uses short, natural probes that real people respond to and clones don't.
CRM, sales tools, helpdesks, telephony, meetings, warehouses, BI, no-code and AI frameworks — if it can make an HTTPS call, it can use Sentient.
Copy-paste REST, client tokens for web and mobile, and drop-in tool definitions so any LLM agent can read emotion and remember how people feel.
curl https://sandbox.api.sentientmachines.tech/v1/analyses/text \ -H "Authorization: Bearer $SENTIENT_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "mode": "sync", "format": "chat", "messages": [ {"speaker": "customer", "text": "Third time I have called about this."}, {"speaker": "agent", "text": "I am sorry. I will fix it now."} ], "reference": {"customer_ref": "CUST-2231"} }'
import os, requests r = requests.post( "https://sandbox.api.sentientmachines.tech/v1/analyses/text", headers={"Authorization": f"Bearer {os.environ['SENTIENT_API_KEY']}"}, json={"mode": "sync", "format": "email", "text": "I'm really worried about this bill."}, ) eri = r.json()["summary"]["eri"] print(eri["score"], eri["band"])
const res = await fetch( "https://sandbox.api.sentientmachines.tech/v1/profiles/customers/CUST-2231", { headers: { Authorization: `Bearer ${process.env.SENTIENT_API_KEY}` } } ); const profile = await res.json(); if (profile.deviation.flagged) { queueProactiveContact(profile.customer_ref, profile.recommended_action); }
// Any MCP-capable assistant or agent framework { "mcpServers": { "sentient": { "command": "npx", "args": ["-y", "@sentient/mcp-server"], "env": { "SENTIENT_API_KEY": "snt_live_…" } } } } // Tools: sentient_analyse_text · sentient_get_emotional_profile // sentient_check_voice_authenticity · sentient_team_insights
Emotion data deserves the highest standard of care — for your customers and your colleagues.
Pseudonymised references only. Audio isn't retained unless you choose.
Card numbers, account numbers, phone numbers and emails masked by default.
A restricted permission for supporting people — never pricing, targeting or matching.
EU mode removes voice-based emotion inference on employees. UK and EU regions.
Start free in the sandbox. Scale on pay-as-you-go — or talk to us about enterprise and platform partnerships.