jambonz vs Vonage Voice API + MCP

jambonz

OpenAPI 3.1 and AsyncAPI documents, an llms.txt and Markdown pages cover the REST and WebSocket APIs, and session prices are public. No status page, security.txt or advisory was found, API keys carry no scopes, and the 2021 privacy policy says audio is never stored while the October 2026 release notes say support keeps call audio for a few days.

Vonage Voice API + MCP

Websocket audio as 16-bit linear PCM at 8, 16 or 24 kHz, both directions. First call needs an application, a private key and a signed JWT.

jambonz

1. Send Authorization: Bearer <token> to https://api.jambonz.cloud/v1, the server in the OpenAPI files. The REST introduction page gives https://jambonz.cloud/api instead, so confirm the base URL with the operator
2. Create a call with from, to and either application_sid or call_hook. Do not retry a timed-out create without checking GET /Accounts/{AccountSid}/Calls first
3. Read ratelimit-remaining and ratelimit-reset on every response and stop when a 429 arrives
4. Append #rc=3&rp=5xx,ct to an application URL to turn on webhook retries. HTTP webhooks are not retried by default
5. Check the account has a carrier and speech credentials before dialling. jambonz does not supply numbers, minutes or speech services

Vonage Voice API + MCP

1. Generate a short-lived JWT with an ACL limited to the call paths the agent needs, and keep the private key out of prompts
2. Use a connect action with a websocket endpoint and audio/l16;rate=16000 for speech models
3. Keep the api.nexmo.com hostname, Vonage's llms.txt says not to swap it
4. Enable signed webhooks and verify the HS256 JWT on each event
5. Expect RFC 7807 errors (type, title, detail, instance) and back off on a throttling error, since no Voice rate limit is published