Switch from Vapi, Retell, or Bland in 15 Minutes

If you're here, you've probably already felt the pain points: unpredictable per-minute bills, audio routing through someone else's servers, or bolting together STT + LLM + TTS pipelines just to get a voice agent that still sounds laggy.

TTMA is different. Your agent runs on your own server with native speech-to-speech (no pipeline), fixed monthly pricing, and a full agent brain with memory, skills, and CRM built in. For a detailed breakdown, see the full comparison.

The good news: your existing work is not wasted. Your prompts, call scripts, and API integrations all carry over. The concepts map cleanly, and TTMA's format is often simpler.

• A Linux VPS - any provider (Hetzner, DigitalOcean, Hostinger). $5-10/month, 2 GB RAM minimum.
• 15 minutes - that covers install, prompt porting, and a test call.
• Your existing call scripts/prompts - export your system prompt, agent config, or pathways from your current platform.

SSH into your VPS and run the installer. It sets up the voice gateway, configures a Cloudflare tunnel (no open ports needed), and gets you a working agent in one command.

curl -fsSL https://voice.moltbot.ninja/install | bash

The installer walks you through initial setup interactively. When it finishes, you'll have a running voice gateway ready to accept calls. Visit voice.moltbot.ninja for detailed install docs.

Your agent's personality and call script live in a playbook - a plain-text file that tells the agent how to handle calls. Here is how prompts from each platform map over.

1 From Vapi (system prompt)

// Vapi assistant config (before)
{
  "model": {
    "provider": "openai",
    "systemPrompt": "You are a dental office receptionist named Sarah.
      You help patients schedule appointments.
      Always confirm the date and time before booking."
  }
}

# TTMA playbook (after) - just paste your prompt as plain text
You are a dental office receptionist named Sarah.
You help patients schedule appointments.
Always confirm the date and time before booking.

2 From Retell (agent config)

// Retell agent config (before)
{ "agent_name": "Sarah", "response_engine": { "type": "retell-llm" },
  "general_prompt": "You are a dental office receptionist..." }

# TTMA playbook (after) - same idea, simpler format
You are a dental office receptionist named Sarah.
...your prompt text here...

3 From Bland (Pathways)

Bland Pathways use a visual node graph. In TTMA, the same branching logic goes into your playbook as conditional instructions, without needing explicit graph edges.

# Bland pathway: "Greeting" -> "Qualify" -> "Book" or "Decline"
# becomes playbook instructions like:

Greet the caller warmly and ask how you can help.

If the caller wants to schedule an appointment:
  - Confirm their preferred date and time
  - Use the book_appointment tool to reserve the slot

If the caller is not interested:
  - Thank them for calling and end the conversation politely

If your current platform calls external APIs (booking systems, CRMs, databases), those integrations map to TTMA's voice-tools.json file. Each tool is a JSON object describing the API call and when the agent should use it.

// voice-tools.json - a tool your agent can call mid-conversation
[
  {
    "name": "book_appointment",
    "description": "Book a dental appointment for the caller",
    "endpoint": "https://your-api.com/appointments",
    "method": "POST",
    "parameters": {
      "patient_name": { "type": "string", "required": true },
      "date": { "type": "string", "required": true },
      "time": { "type": "string", "required": true }
    }
  }
]

If you had Vapi "Functions" or Retell "Custom LLM" webhooks, the same logic moves here. The endpoint, method, and parameters translate directly.

TTMA uses a dedicated telephony provider for phone connectivity. You have two options:

• Port your existing number - submit a porting request through the telephony provider's dashboard. Porting typically takes 1-2 weeks depending on your current carrier. Your old platform stays active during the transition.
• Get a new number - instantly provision a new number through your TTMA dashboard. You can start testing immediately while the port processes in parallel.

Either way, once the number is connected to your TTMA instance, incoming calls route directly to your voice gateway.

Place a test call to your TTMA number. Verify that:

• The agent greets you as expected (playbook is loaded)
• Tool calls work (try triggering a booking or lookup)
• Barge-in works (interrupt mid-sentence and confirm it responds naturally)
• The call log shows up in your dashboard with transcript and recording

If something sounds off, edit the playbook and call again. Changes apply immediately - no redeploy required.

Use this table as a quick reference when translating your setup.