Add Tools and Human Transfer to a Vonage + Deepgram Voice Agent

A Vonage tutorial shows how to extend a Deepgram-powered phone voice agent beyond conversation by adding a controlled order-status lookup tool, strict latency limits, deterministic failure handling, human transfer, and call logging. The Node.js example registers a client-side Deepgram function that lets the application, rather than the model, authorize and execute a mocked order lookup, allowing only one request per call and retrying only temporary transport failures. To avoid silence during live calls, lookups time out after 1.5 seconds and trigger a fixed spoken fallback, while unknown orders remain valid lookup results. Requests involving billing, returns, cancellations, or account issues are identified through predefined keywords and transferred through Vonage to a support number using an NCCO. Each interaction is saved to SQLite with transcripts, tool activity, latency, fallback and handoff details, outcomes, and agent version information, enabling teams to identify recurring problems and evaluate future capabilities.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.