Create a ThunderPhone agent

1. Select a tier. Use current prices from the pricing guide, not cached estimates.

2. Resolve real options. Call GET /v1/voices and use its returned voice ID. Check the supported-languages guide before choosing language. Do not invent a voice ID, locale, or product string.
3. Write the prompt with thunderphone-prompt-builder. Keep policy and facts general; do not script caller-response branches.
4. Create the agent. name, prompt, and voice are required. The create response is deployed revision 1. Valid product values are listed in references/agent-fields.md.

{
  "name": "Dental receptionist",
  "prompt": "You are the receptionist for the dental practice. Help callers with scheduling and practice information. Use the caller's language when supported. Keep spoken responses concise.",
  "voice": "${VOICE_ID}",
  "product": "bolt",
  "primary_language": "en-US"
}

curl --fail-with-body -X POST https://api.thunderphone.com/v1/agents \
  -H "Authorization: Bearer $THUNDERPHONE_API_KEY" \
  -H "Content-Type: application/json" \
  -H "X-ThunderPhone-Client: skills/create-agent@1.0" \
  --data-binary @agent.json

import os, requests

payload = {"name": "Dental receptionist", "prompt": os.environ["AGENT_PROMPT"],
           "voice": os.environ["VOICE_ID"], "product": "bolt", "primary_language": "en-US"}
r = requests.post("https://api.thunderphone.com/v1/agents", json=payload,
    headers={"Authorization": f"Bearer {os.environ['THUNDERPHONE_API_KEY']}",
             "X-ThunderPhone-Client": "skills/create-agent@1.0"}, timeout=30)
r.raise_for_status()
agent = r.json()
print(agent["id"])

const response = await fetch("https://api.thunderphone.com/v1/agents", {
  method: "POST",
  headers: { Authorization: `Bearer ${process.env.THUNDERPHONE_API_KEY}`,
    "Content-Type": "application/json", "X-ThunderPhone-Client": "skills/create-agent@1.0" },
  body: JSON.stringify({ name: "Dental receptionist", prompt: process.env.AGENT_PROMPT,
    voice: process.env.VOICE_ID, product: "bolt", primary_language: "en-US" }),
});
if (!response.ok) throw new Error(await response.text());
console.log((await response.json()).id);

5. Edit deliberately. PATCH /v1/agents/{agent_id} creates or updates a draft. Inspect it, then POST /v1/agents/{agent_id}/deploy. Use POST /v1/agents/{agent_id}/discard-draft to abandon unshipped changes.
6. Verify. Fetch GET /v1/agents/{agent_id} and confirm the deployed and draft state matches the intended revision. Then run the test-agent skill.

Errors

Source and safety rules

• Confirm fields in the agents API reference.
• Confirm tier rates in pricing.
• Use supported languages and the live voices endpoint.
• Never fabricate IDs, credentials, voices, prices, or successful deployment evidence.
• Treat deployment as a behavior change; test the exact deployed revision.