Free
Evaluation and prototyping.
$0forever
- Voice minutes / mo100
- Active agents1
- Seats1
- Trace retention7 days
Pricing that scales with you.
Free for evaluation, Hobby for side projects, Pro for production, Team for growth. Every tier ships the complete memory engine — no per-operation metering, no hidden quotas. Bring your own STT, LLM, and TTS keys; Hyponema never bills provider minutes.
Hobby
Side projects and small launches.
$29/ month
- Voice minutes / mo1,000
- Active agents5
- Seats3
- Trace retention30 days
Pro
Production applications.
$199/ month
- Voice minutes / mo10,000
- Active agents30
- Seats10
- Trace retention90 days
Team
Growing teams with scale needs.
$599/ month
- Voice minutes / mo30,000
- Active agents100
- Seats30
- Trace retention1 year
- 1 agent · 1 seat · 100 voice minutes / month
- Full relational memory engine — importance scoring, narrative arcs, persona drift, session resume
- Persona builder with identity, personality scales, no-go zones, sleep windows
- All STT / LLM / TTS providers. Bring your own keys, unlimited API keys on every plan
- GDPR DSAR export · audit log · fatigue detection
- Community support
- 5 agents · 3 seats · 1,000 voice minutes
- Knowledge base / RAG — up to 10 documents, hybrid semantic + lexical retrieval
- MCP server — plug the memory engine into Claude Desktop, Vapi, Retell, ElevenLabs
- Audio downloads of every conversation
- Email support · 24h response on weekdays
- + Everything in Free
- 30 agents · 10 seats · 10,000 voice minutes
- Knowledge base / RAG — up to 100 documents
- Workspace sharing with role-based access (Owner, Admin, Designer, Developer, Viewer)
- Scenario tests, tool-call tests, simulation tests with LLM-judge scoring
- Priority email support · 24h response
- + Everything in Hobby
- 100 agents · 30 seats · 30,000 voice minutes
- Knowledge base / RAG — up to 500 documents
- SSO / SAML for your identity provider
- Slack support channel with the Hyponema team
- Everything in Pro
- + Everything in Pro
Every tier ships the complete memory engine — importance scoring, narrative arcs, persona drift, emotional trajectory, supersede chains, session resume. No per-operation metering, no hidden memory quotas. Unlimited API keys on every plan.
Prices in USD. 15% discount on annual billing. Bring your own STT / LLM / TTS keys — Hyponema does not bill provider minutes. Enterprise plans with custom limits and dedicated infrastructure available on request.
Estimate
What will you actually pay?
Move the sliders. We point you to the right tier. STT, LLM, and TTS are billed by your provider — Hyponema never marks them up.
Voice minutes per month2,000 min
Active agents3 agents
Seats3 seats
Every plan runs on your own STT, LLM, and TTS API keys. Unlimited API keys on all tiers. Memory, tools, and evaluations are never metered — they are included transparently.
Suggested tier
Pro
- Hyponema subscription
- $199
- Provider cost (your keys)
- billed by your provider
Compare
Everything by tier.
| Free | Hobby | Pro | Team | |
|---|---|---|---|---|
| Voice minutes / month | 100 | 1,000 | 10,000 | 30,000 |
| Active agents | 1 | 5 | 30 | 100 |
| Seats | 1 | 3 | 10 | 30 |
| Trace retention | 7 days | 30 days | 90 days | 1 year |
| Knowledge base documents | 1 | 10 | 100 | 500 |
| Memory engine | ||||
| Persona drift detection | ||||
| Unlimited API keys | ||||
| MCP server | ||||
| GDPR DSAR export | ||||
| Audit log | ||||
| Sleep windows + fatigue detection | ||||
| SSO / SAML | ||||
| Support | Community | Priority email | Slack |
Questions
Things people ask before signing up.
Product
Can I really use any STT, LLM, and TTS combo?
Yes. Three pre-validated combos (Fastest / Balanced / Cheapest) plus any custom mix across Deepgram, AssemblyAI, Whisper, Anthropic, OpenAI, Gemini, Cartesia, ElevenLabs, OpenAI TTS. Cascading is independent per layer with up to three retries. Custom OpenAI-compatible LLMs plug in but don't participate in the cascade.
What does "relational memory" actually mean?
Narrative arcs (ACTIVE / DORMANT / RECURRING) detected across weeks. Emotional trajectory: 14-day rolling vs 60-day baseline drift, computed without an LLM call. Persona-consistency check pre-emit on every assistant turn. Topic cooldowns auto-suppress sensitive users for 3 days after a NEGATIVE / CONCERNED episode. Supersede chains for contradictions — never silent overwrites.
How fast is time-to-first-conversation?
Under five minutes for the median user. Signup is magic-link only — no password, no credit card. The dashboard ships seed data and a "Test Agent" button that opens a real WebRTC call directly in the browser.
Vendors & data
Where do my provider API keys live?
AES-256-GCM envelope encryption: a per-credential DEK wrapped by a KMS-managed KEK. Decrypted only in-process for the seconds a call lasts. Operators on the Hyponema side never see them in plaintext. (Conversations themselves are not customer-key-encrypted — they are tenant-isolated by Postgres RLS and surfaced only to your workspace members.)
Can I export everything if I leave?
Yes. The DSAR endpoint zips every user, conversation, memory, narrative arc, audit log, and configuration as JSON, plus audio recordings when configured. One endpoint, async job, signed download URL. The "forget" cascade does the inverse: every memory, arc, thread, embedding, transcript, and audio file purged with an audit trail of the deletion.
Compliance & ethics
Are you SOC 2 / HIPAA compliant?
GDPR DPA, CCPA, and DSAR export ship on every plan today. SOC 2 Type II, HIPAA BAA, and SSO are on the roadmap. Talk to us if you need a timeline or a roadmap letter for procurement.
What about parasocial harm?
Sleep windows (user-local quiet hours with overnight + weekday support), fatigue detection (≥ 5 sessions > 30 min in 24 h, or ≥ 90 min total, or 3 consecutive NEGATIVE / CONCERNED sentiments), and topic cooldowns are built in and on by default on every plan. Tunable per persona.
Surfaces
Where does Hyponema run?
Three surfaces: the dashboard at app.hyponema.ai, a public REST + WebSocket API (40+ routers, OpenAPI 3.1, RTVI over Pipecat for voice), and an embeddable widget — Preact + Vite IIFE under 50 KB gzipped — that drops into any page with one line of HTML. Native iOS / Android / Python SDKs are post-MVP; today the WS session protocol works from any duplex client.
How do I take it to phone?
Telnyx and Twilio ship today, end-to-end on both control plane and media plane (inbound and outbound). Vendor-neutral by construction — Vonage / SignalWire / direct-SIP slot in by mirroring the Twilio file structure. Hyponema also exposes the MCP server (§5.16 in the docs) so customers staying on ElevenLabs / Vapi / Retell can plug the memory engine in via standard MCP tool calls.
Can I use it from React Native or embedded hardware?
Native mobile and hardware SDKs are post-MVP. Today the WebSocket session protocol works from any duplex client; embedding into a React Native app or a custom device today means writing a thin client against the same WS spec the widget uses.
Voice agents built for years, not minutes.
Bring your own keys. Join the waitlist for early access.