This April, we focused on shipping tools to help you connect Vapi agents more seamlessly, support early-stage builders, and refine the core platform based on your feedback.
Here’s everything we shipped this month:
🚀 New Feature
We tackled common integration and communication challenges with these additions:
- Vapi MCP Client
Vapi now natively supports the Model Context Protocol (MCP). Point your Vapi agent to any MCP-compatible server (like Zapier Actions or Composio) and instantly unlock thousands of pre-built actions across countless apps.
Learn how to configure it in the Vapi MCP Client Docs.
- Vapi MCP Server
Now you can trigger Vapi actions directly from MCP-compatible clients like Claude Desktop. Use natural language in your chat interface to bridge the gap between text-based AI and real-world phone infrastructure. Get started with the Vapi MCP Server Docs.
- Vapi SMS
Voice agents can now send and receive SMS messages directly within a Vapi session (currently via Twilio). This enables your agents to switch seamlessly between voice and text based on user interaction and support fully text-based flows when needed. See the SMS Setup Guide.
🤝New Program
Announcing the Vapi Startups Program
Building voice AI from the ground up takes resources. If you're an early-stage founder building with voice AI, we want to help. The new Vapi Startups Program offers:
- 90,000 free minutes to prototype and test
- Priority access to new features
- Private Slack community with technical support and feedback sessions
Eligibility criteria: Your startup must be founded <5 years ago, funded $100K-$5M, and be actively building with voice AI.
Apply to the Vapi Startups Program here
🤖New Models and Voices
We added more options to fine-tune performance, accuracy, and personality:
- New LLMs: Added support for o3-mini, o4-mini, and GPT 4.1 models.
- Enhanced Transcription: Deepgram Nova-3 now supported, with stronger accuracy on multilingual and medical use cases.
- New Voices: You can now use Sesame (update: depreciated now) and Deepgram Aura-2 voices into your assistants.
Check your Vapi Dashboard to explore these options.
🧰 Platform Enhancements
We’ve made it easier to build and debug large-scale voice systems:
- Pre-Answer Call Hooks: Configure
say or transfer actions directly on phone numbers to disable inbound calling [Docs]
- Batch Calls & Scheduling: Initiate multiple outbound calls efficiently via API array (
/call/batch) and schedule calls within specific time windows using SchedulePlan. [Docs]
- New Built-in Tools: Easily integrate with Slack, Google Calendar, & Google Sheets using new pre-built tools (requires OAuth setup). [Docs]
- Multi-Structured Data Extraction: Define multiple schemas (
structuredDataMultiPlan) to run multiple structured data analysis plans at the end of the call. [Docs]
- WebSocket Call Support (
vapi.websocketCall): Establish calls over WebSockets for alternative connectivity, direct web app integration, and potential latency improvements. [Docs]
What are the next big hurdles you're facing in production or development? Where should we focus our efforts next?
Let us know in the Discord or here.
As always, we're eager to see what you build.
