FAQ
Answers to the most common questions about Neuratel. Can't find what you need? Email [email protected].
Getting Started
How do I get access to Neuratel?
Neuratel is invite-only with a hands-on onboarding process. Start with a free technical consultation at neuratel.ai where we analyze your stack and operations. Once aligned, we set up your entire project within 24 hours — agents configured, numbers connected, integrations wired — then send your platform invite. See Account Setup for the full process.
What languages are supported?
Neuratel supports 30+ languages through its provider integrations. Commonly used languages include English (US, UK, AU), Arabic, Spanish, French, German, Portuguese, Italian, Japanese, Chinese (Mandarin), and Hindi. Language support depends on the STT and TTS providers configured for your agent.
Where do I see usage and balance?
From Organization → Billing in the dashboard, or via the API — see Usage Visibility. Pricing, contract terms, and any change to billing setup go through your account team.
Can I use my own API keys for providers?
Yes. Neuratel supports BYOK (Bring Your Own Key) for major providers. Contact your account team to enable it.
Agents
What is Phantom Brain?
Phantom Brain is Neuratel's proprietary LLM optimized specifically for real-time voice conversations. It is designed for ultra-low latency responses, natural turn-taking, and voice-specific interaction patterns. Available as a brain provider option alongside OpenAI, xAI (Grok), and Groq.
Can I run multiple agents on the same phone number?
No. Each phone number is assigned to exactly one agent at a time. To serve different use cases, provision separate numbers for each agent. You can route multiple local numbers from different area codes to the same agent.
What is the maximum call duration?
Call duration is configurable per agent via the conversation.max_duration setting. There is no hard platform limit. Most customers set durations between 10 and 60 minutes depending on the use case. If no max is set, calls continue until the caller or agent hangs up or the silence timeout triggers.
How do I make my agent sound natural?
- Keep responses short: 1-2 sentences per turn
- Use conversational language in the system prompt, not corporate jargon
- Choose a voice that matches the agent's persona
- Enable backchannel responses ("mm-hmm", "I see")
- Adjust TTS
stabilityandsimilarity_boostto find the right balance - Test with real callers and iterate on the system prompt
Phone Numbers
How do I get phone numbers?
Your Neuratel account team provisions numbers for you. Contact support to request numbers for your region (local or toll-free). You can also port your existing numbers. See Setting Up Phone Numbers.
Can I port my existing numbers?
Yes. Contact [email protected] with the numbers to port and your current carrier information. The process takes 3-7 business days with no downtime.
Calls
What is the difference between blind and warm transfer?
Blind transfer: The agent connects the caller directly to the destination and disconnects immediately. No introduction or context is passed to the human.
Warm transfer: The agent stays on the line, introduces the caller to the human operator, provides context, and then disconnects. Better for complex situations where context matters.
Can I record calls?
Yes. Call recording is configurable per agent. When enabled, recordings are stored securely and accessible via the API and dashboard. Note that many jurisdictions require consent disclosure — configure your agent's greeting to include a recording notice if applicable.
How do I handle concurrency?
Check current concurrent call count:
Code
Response includes active_calls count and your plan's max_concurrent limit. For outbound campaigns, set max_concurrent_calls on the campaign to stay within your limit.
Knowledge Base
What file formats does the knowledge base accept?
- PDF — up to 21 MB per file
- DOCX — up to 21 MB per file
- TXT — up to 21 MB per file
- URL — web pages are crawled and indexed
Upload via the dashboard or API. Documents are chunked, embedded, and made available for RAG (Retrieval-Augmented Generation) during calls.
How does the agent use the knowledge base?
When a caller asks a question, the agent searches the knowledge base for relevant chunks of text, then uses that context to generate an accurate, grounded response. This is called RAG. The agent cites knowledge base content rather than hallucinating answers.
Webhooks
How do I test my webhook endpoint?
Send a test event using the API:
Code
This sends a test payload to your endpoint. Check the delivery logs to confirm receipt:
Code
What events can I subscribe to?
Currently available: call.started, call.ended, and call.failed. Additional events (recording.ready, call.summary.ready, transcript.*, agent.*) are defined and coming soon. See Webhook Events for the full list and availability.
My webhook is not receiving events. What should I check?
- Webhook URL is publicly accessible from the internet
- URL uses
https://(HTTP is not accepted) - Your server returns 200 within the timeout window (default 30 seconds)
- Firewall or security group allows inbound POST requests from Neuratel
- Webhook is set to
is_active: true - You are subscribed to the correct event types
See Troubleshooting for more.
API
What API key formats exist?
All Neuratel API keys start with nk_live_.
How do I rotate my API key without downtime?
Use the Rotate button in Settings → API Keys. The old key stays valid for a configurable grace period while you deploy the new one, then expires automatically. See Using API Keys.
Providers
Which LLM providers are supported?
OpenAI, xAI (Grok), Groq, and Phantom Brain (Neuratel's proprietary model). Each can be configured as the agent's brain provider with BYOK credentials.
Which voice (TTS) providers are supported?
ElevenLabs, Cartesia, and Phantom Speech. Cartesia offers the lowest latency (~40ms TTFA), while ElevenLabs offers the most expressive and customizable voices. Phantom Speech is Neuratel's proprietary TTS optimized for voice conversations.
Which transcription (STT) providers are supported?
Deepgram Nova-3 and OpenAI gpt-4o-mini-transcribe. Phantom STT is coming soon.
Support
How do I contact support?
- Email: [email protected]
- Slack: Dedicated workspace for enterprise clients
- Status page: status.neuratel.ai
Is there an SLA?
Enterprise accounts include SLA guarantees. Contact [email protected] for details.
Next Steps
- Troubleshooting — detailed solutions to common issues
- Glossary — terminology reference
- Production Checklist — pre-launch verification