Voice CallSAFE
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
A Model Context Protocol (MCP) server that enables Claude and other AI assistants to initiate and manage voice calls using Twilio and OpenAI (GPT-4o Realtime model).
Use this as a base to kick-start your AI-powered voice calling explorations, save time and develop additional functionality on top of it.
Sequence Diagram
sequenceDiagram participant AI as AI Assistant (e.g., Claude) participant MCP as MCP Server participant Twilio as Twilio participant Phone as Destination Phone participant OpenAI as OpenAI AI->>MCP: 1) Initiate outbound call request (POST /calls) MCP->>Twilio: 2) Place outbound call via Twilio API Twilio->>Phone: 3) Ring the destination phone Twilio->>MCP: 4) Call status updates & audio callbacks (webhooks) MCP->>OpenAI: 5) Forward real-time audio to OpenaAI's realtime model OpenAI->>MCP: 6) Return voice stream MCP->>Twilio: 7) Send voice stream Twilio->>Phone: 8) Forward voice stream Note over Phone: Two-way conversation continues until the call ends
Features
- Make outbound phone calls via Twilio 📞
- Process call audio in real-time with GPT-4o Realtime model 🎙️
- Real-time language switching during calls 🌐
- Pre-built prompts for common calling scenarios (like restaurant reservations) 🍽️
- Automatic public URL tunneling with ngrok 🔄
- Secure handling of credentials 🔒
Why MCP?
The Model Context Protocol (MCP) bridges the gap between AI assistants and real-world actions. By implementing MCP, this server allows AI models like Claude to:
- Initiate actual phone calls on behalf of users
- Process and respond to real-time audio conversations
- Execute complex tasks requiring voice communication
This open-source implementation provides transparency and customizability, allowing developers to extend functionality while maintaining control over their data and privacy.
Requirements
- Node.js >= 22
80925bef48e0OBSERVED · 2026-10-07Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add voice-call-mcp-server --env OPENAI_API_KEY=${OPENAI_API_KEY} --env TWILIO_AUTH_TOKEN=${TWILIO_AUTH_TOKEN} -- npx -y voice-call-mcp-server{
"mcpServers": {
"voice-call-mcp-server": {
"command": "npx",
"args": [
"-y",
"voice-call-mcp-server"
],
"env": {
"OPENAI_API_KEY": "${OPENAI_API_KEY}",
"TWILIO_AUTH_TOKEN": "${TWILIO_AUTH_TOKEN}"
}
}
}
}Exposed tools (1)
0 read · 1 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
trigger-call | write | Trigger an outbound phone call via Twilio |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (5 observation(s))
- Network
- declared (5 observation(s))
- Shell
- none-observed
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (7)
import { generateOutboundCallContext } from '../../config/prompts.js';import { CallState, ConversationMessage } from '../../types.js';import { CallState } from '../../types.js';import { LOG_EVENT_TYPES, SHOW_TIMING_MATH } from '../../config/constants.js';import { checkForGoodbye } from '../../utils/call-utils.js';@ngrok/ngrok, axios, body-parser, colors, cors, dotenv, eslint-plugin-simple-import-sort, express
assets/demo.gif
Gates applied: no_behavioural_pass.
80925bef48e0full audit observations/trust-audit/mcp-server/popcornspace__voice-call.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-07 | 80925bef48e0 | SAFE | B | 89 | first audit |
Questions
What tools does Voice Call expose?
1 in total: 0 read-only, 1 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is Voice Call safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B.
What credentials does Voice Call need?
It reads OPENAI_API_KEY and TWILIO_AUTH_TOKEN from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does Voice Call run?
It speaks stdio, so it runs as a local process your client starts. It is published on npm as voice-call-mcp-server.
How current is this page?
The grade is for one exact copy of the source (80925bef48e0), read on 2026-10-07. The repository is watched and re-audited when it changes.