Talknex is an enterprise-grade platform for deploying conversational AI phone agents. You configure an agent with a persona, voice, and system prompt; connect a phone number via Twilio; and the agent autonomously handles inbound and outbound calls — booking appointments, qualifying leads, and transferring to humans when needed.
No. The full dashboard — agent configuration, knowledge base uploads, call analytics, and phone number management — is entirely no-code. Developers who want to go deeper can use our REST API, WebSocket streaming endpoints, and webhooks.
Most teams are live in under 15 minutes. Sign up → create an agent (paste a system prompt and pick a voice) → provision a phone number → your agent starts answering calls. No infrastructure to manage.
Yes. Every agent has a built-in browser-based call tester so you can run a full conversation before attaching a real number. You can also use our REST API to simulate calls programmatically in your CI pipeline.
Yes. You can trigger outbound calls via the dashboard or API — useful for follow-ups, appointment reminders, or lead qualification campaigns. Outbound minutes are billed the same as inbound minutes.
Talknex agents can understand and speak 30+ languages out of the box through our Deepgram STT and ElevenLabs TTS integrations, including English, Spanish, French, German, Portuguese, Arabic, Hindi, Mandarin, and many more. Custom language models are available on Fully Managed.
No. We pipeline three leading AI providers — Deepgram for streaming speech-to-text, GPT-4o for reasoning, and ElevenLabs Turbo v2 for text-to-speech — so the agent responds with natural prosody and emotion. Most callers cannot distinguish it from a human within the first minute.
Our median end-to-end round-trip is ~580ms — consistently under 600ms. We achieve this by pipelining streaming STT, LLM, and TTS: audio starts playing before the LLM finishes generating.
Yes. Deepgram VAD (voice activity detection) detects when the caller speaks over the agent and immediately halts TTS playback. You can tune interruption sensitivity per-agent — from highly interruptible (conversational style) to low sensitivity (good for complex explanations).
Yes — voice cloning is available via ElevenLabs on Fully Managed plans. Provide ~10 minutes of clean audio and you'll have a custom voice model attached to your agent, so you can ship your brand voice at scale.
Calls run over Twilio's carrier-grade PSTN infrastructure. If the WebSocket session drops, the call gracefully disconnects and the partial transcript is flushed to the database immediately. The agent will not attempt to reconnect mid-call — the caller must dial back.
Pricing is usage-based, not tiered. Bring Your Own Keys (BYOK) starts at $0.05/min with no base fee, or go Fully Managed from a $99/mo base plus usage. You only pay for the call minutes and concurrent lines you actually use — estimate your bill on the pricing page. Secure checkout is powered by LemonSqueezy, so there are no surprise invoices.
No. On Bring Your Own Keys (BYOK) there is no base fee and no minimum — you pay only for usage, starting at $0.05/min, and your first concurrent line is free. There is no fixed monthly tier to commit to; you can start self-serve and scale as your call volume grows.
Yes. Because billing is usage-based, you can set a hard spend cap in the dashboard. You are never cut off mid-call — but once the cap is hit, new inbound calls receive a polite voicemail message and are logged for follow-up, so your bill stays predictable.
Yes. With Bring Your Own Keys (BYOK) you can connect your own Twilio SID and Auth Token. Twilio usage will then bill directly to your Twilio account — Talknex only bills its platform usage fee. This is useful for teams with existing Twilio number inventories.
Billing is usage-based and charged monthly for the minutes and concurrent lines you actually used. There is no annual contract or lock-in — you can cancel anytime, and on BYOK there is no minimum commitment.
Native integrations exist for Salesforce and HubSpot. After each call, transcripts, sentiment scores, and call outcomes are synced as activities. Contact and lead records are created automatically if the caller's number is new. You can connect other tools via Zapier, Make, or webhooks.
Yes — the built-in book_appointment tool writes directly to Google Calendar or Microsoft Outlook (via OAuth). The agent checks your real availability, proposes slots, confirms with the caller, and sends an invite with a meeting link. No back-and-forth email required.
Yes. There are native Zapier and Make triggers for call completed, new lead created, and appointment booked events. You can chain these to 5,000+ apps without writing code — Slack alerts, Airtable rows, Google Sheets, and more.
Calls run over Twilio's carrier-grade PSTN infrastructure. On Bring Your Own Keys (BYOK) you can connect your own Twilio account and number inventory; the same agent config handles inbound and outbound calls, and the voice pipeline adapts automatically.
All data is encrypted in transit with TLS 1.3 and at rest with AES-256. Transcripts, recordings, and vector embeddings are strictly scoped to your organization — no data crosses org boundaries. We do not use customer data to train shared AI models.
Talknex is architected to be HIPAA-ready and we sign a Business Associate Agreement (BAA) with Fully Managed customers. If you are in healthcare, contact our sales team to discuss your specific PHI handling requirements before going live.
By default data is stored in the EU (Frankfurt). Fully Managed customers can request US, EU, or APAC data residency at the org level. We are SOC 2 Type II in progress and GDPR-compliant for EU data subjects.
Yes — recordings and transcripts are your data. You can delete individual call records from the dashboard or bulk-delete via the API. Deletion is permanent and propagates to all storage layers within 24 hours.
A standard REST API under https://api.talknex.ai/v1. Auth is via Bearer tokens. Every resource — agents, calls, contacts, knowledge bases — has full CRUD endpoints with OpenAPI 3.0 docs available at /docs.
Yes — subscribe to a call's WebSocket stream and receive transcript deltas, sentiment events, and tool calls in real-time as the conversation happens. This is how our own dashboard renders the live call widget.
Configure a webhook URL in the dashboard and we POST signed JSON payloads to it on events: call.started, call.ended, transcript.final, appointment.booked, lead.created. Signatures use HMAC-SHA256 with your webhook secret so you can verify authenticity.
Yes. Define custom tool schemas (JSON Schema format) in the agent's Developer tab. When the LLM decides to call a tool, Talknex POSTs the arguments to your webhook URL, waits for your response, and feeds the result back into the conversation — all in under 300ms round-trip.
On Fully Managed, sales-assisted plans add dedicated infrastructure (no noisy neighbors), white-label capabilities, SSO/SAML, custom voice models, a dedicated success manager, SLA-backed support, a HIPAA BAA, and optional on-premise deployment. Contact sales to provision the configuration you need.
Yes — on-premise and private cloud (AWS, GCP, Azure) deployments are available on Fully Managed, sales-assisted plans. We provide a Docker-based bundle with the full voice stack. Your data never leaves your network.
Yes. White-label removes all Talknex branding, allows custom domain and logo, and optionally white-labels the email sender domain. Resellers and agencies use this to power their own AI voice products.
Still have questions?
Our team is available to walk you through onboarding, integrations, or compliance requirements.