Vapi

Testing voice agents is hard. Really hard. LLMs are already unpredictable. Now throw in real-time audio, barge-ins, streaming APIs, long context windows, RAG, and phone latency. That’s what testing Voice AI looks like. We’ve seen it all at Vapi: - Agents that talk over users - Context out of thin air - Tool-calls that fire twice or not at all And you can’t debug all this with transcript analysis. Every conversation is a messy, timing-sensitive dance. So we built something better. 🎯 Vapi Test Suite: a purpose-built way to test real-time voice agents with expected inputs, agent behaviors, interrupt logic, and detailed call traces. - Define a set of test cases for agents - Simulate calls with voice or text input - Catch regressions before users do If you’re building voice agents, you know testing is the difference between “cool demo” and “production ready.” Give it a try and let us know what you think.

Much vibe-coding, real-coding, late nights, and blue bottle runs later...we launched Vapi 2.0. LFG Vapi 🥳

Vapi 2.0 is live! 🚀 We help developers deploy real-time voice agents - LLM-powered, streaming, and production-ready. Vapi 2.0 introduces: - 100+ model integrations (transcription, LLM, TTS) - Conversation workflows and knowledge bases - Vapi voices that sound human - Rock-solid enterprise infrastructure to scale to millions of calls - Real-time observability with Langfuse integration - A metric ton of voice agent configuration options We’ve powered 44M+ calls across healthcare, support, and logistics. Voice infra shouldn't take months to build and it shouldn’t break under load. 📍 https://vapi.sh/4G6m0MG 📞 Call it live: 1-844-HEY-VAPI PS: Team Vapi has hidden Easter eggs on their Product Hunt page 👀 Spot them and drop a comment on the PH page to claim your share of the $10,000 prize pool 🥚💰

We're #hiring a new Solutions Engineer in San Francisco, California. Apply today or share this post with your network.

No Calendly links. No timezone math. No scheduling chaos. Just one phone number… and 20 participants calling in on their own time. Christian Ulstrup —founder of GSD at Work LLC used Vapi to completely transform how he preps for his high-impact AI workshops. He spun up a voice agent with custom prompts, dropped it into a workshop invite, and let it run. - Each participant called in whenever it worked for them - Vapi captured rich, unscripted context - Transcripts + summaries landed directly in Slack + HubSpot - And the workshop? Highest-rated of the entire event That’s the power of voice agents that actually do stuff. Want the full breakdown with the complete script of the agent he built? Video + workflow here: https://vapi.sh/p3yATu4

Most “voice agents” are just fancy IVR scripts. At Vapi, we’re building actual agents that can: - Access live data in real-time - Make logic-based decisions - Stay stateless, but behave statefully To make that work, we needed a database that wouldn’t slow us down. We use Neon because it supports: - Autoscaling for spiky workloads - Reliable connection pooling - Streaming replication to ClickHouse via PeerDB It’s one piece of a much bigger system, but a critical one. If you’re curious how we wired it all together, we shared more here:

Our growth team was digging into what it takes to win a Golden Kitty. Only to realize we’d already won one 🏆😂 Vapi won the Golden Kitty for Best AI Voice Product of the Year 2024. Thanks to everyone building with us early. It’s nice to know we’re heading in the right direction. We have just gotten started…

Sesame has been making some big waves in voice AI this past week. Traditional voice agents turn speech into text, put it through an LLM, then convert text back to speech. At each step, all the nuance is lost. Sesame changes this. It’s a speech-to-speech model, which means it can natively hear and produce speech, just like humans do. To help you build faster, we are introducing the Vapi x Sesame integration: - Configure agents with all the usual Vapi configuration and tooling - Deploy Sesame voice agents in phone calls and websites - Test, iterate, and redeploy, all in Vapi. The current release from Sesame, CSM1-B, is only a voice model. We've been waiting a long time for speech-to-speech models to reach human performance, and these are still early steps. Integrate Sesame today: https://vapi.sh/sesame Stay tuned for more updates tomorrow! 🚀

We're #hiring a new SRE in San Francisco, California. Apply today or share this post with your network.