Cipher
HomeVoice ProxyHow It WorksFeaturesDevelopersFAQ
Cipher
Sign In
Proxy PlatformMCP Platform
GET STARTED
  • Overview
INTEGRATIONS
  • Vapi Setup
  • ElevenLabs Setup

MCP Platform

Bring Cipher's trust verification engine to the Model Context Protocol. Enable MCP-compatible assistants to verify caller identity, score trust, and enforce security policies during conversations.

The MCP Platform exposes Cipher's same powerful verification engine used by the Proxy Platform as a set of tools and resources that AI assistants can invoke directly. Perfect for desktop agents, chatbots, and custom AI applications that need caller verification.

What is Model Context Protocol?

Understanding MCP and its role in AI applications

Model Context Protocol (MCP) is an open standard that enables AI assistants to securely connect to external tools and data sources. Think of it as an API specifically designed for AI agents—rather than making HTTP requests, the AI can call tools and read resources to extend its capabilities.

🛠️ MCP Tools

Functions the AI can call to perform actions (e.g., verify_caller to check identity)

📚 MCP Resources

Data the AI can read for context (e.g., session://call-123 to view verification state)

🔌 MCP Servers

Backend services that provide tools and resources (like Cipher's verification server)

How Cipher's MCP Server works

Understanding the verification flow

1

Your MCP client connects to Cipher's MCP server via STDIO or SSE transport. The server registers available tools and resources. You can setup MCP tool calling easily in providers such as ElevenLabs or Vapi or use a custom MCP client if you prefer.

2

During a conversation, Claude can invoke verify_caller to process user messages and extract identity information (name, address, phone). The same NER engine from the Proxy Platform is used.

3

Cipher calculates a trust score (0-100%) based on collected evidence and optionally verifies phone numbers via Twilio. All metrics are persisted to your dashboard for audit trails.

4

Your MCP client receives the verification result and can use it to gate access, personalize responses, or enforce security policies (e.g., "Only proceed if trust score > 80%").

MCP Tools

Functions available for AI assistants to call

verify_caller

Primary Tool

Process a caller's message to extract identity information, update session state, calculate trust score, and generate the next verification question. This is the core verification tool.

Arguments: session_id, caller_message, office_number (optional), conversation_history (optional)
Returns: Trust score, entities detected, next prompt, verification status, collected info

get_session_state

Retrieve current session state including trust score, collected information, evidence, and missing verification factors.

Returns: session data, trust score, evidence list

check_verification_status

Quick check if a caller is verified (trust score ≥ threshold). Useful for gating access to sensitive operations.

Returns: verified boolean, trust score, threshold

verify_phone_number

Explicitly verify a phone number using Twilio caller ID lookup. Cross-references caller name, carrier, and geolocation.

Returns: caller name, carrier info, trust score delta

MCP Resources

Data available for AI assistants to read

session://{session_id}

Real-time session state formatted as markdown. Shows verification status, trust score, collected information, evidence, and missing factors. Perfect for Claude to understand conversation context.

Example: session://call-abc123

schema://trust/default

Trust factor schema showing weights, thresholds, and descriptions for each verification factor. Helps Claude understand what information is needed and why.

Threshold: 80% | Factors: Name, Address, Phone, etc.

Key features

What makes Cipher's MCP platform unique

Same Engine, Different Interface

Uses the exact same trust verification engine as the Proxy Platform. All metrics persist to the same dashboard for unified observability.

Claude Desktop Ready

Designed specifically for Claude Desktop integration via STDIO transport. One config file and you're running.

Web Integration via SSE

Also supports SSE transport for web-based integrations. Connect from Vapi, ElevenLabs, or custom applications over HTTPS.

Prompt Injection Detection

Automatically scans incoming messages for prompt injection attacks using ML models. Penalizes trust score if malicious patterns detected.

Dual-platform architecture: Choose the Proxy Platform for voice agent integrations (ElevenLabs/Vapi), or the MCP Platform for desktop AI assistants (Claude Desktop). Both share the same verification engine and dashboard.

Next steps

Vapi Setup →

Integrate Cipher MCP with Vapi assistants

ElevenLabs Setup →

Integrate Cipher MCP with ElevenLabs agents

Proxy Platform →

Integrate Cipher with voice agent platforms instead

Dashboard →

Create API keys and monitor verification activity

2025 Cipher Inc

·Terms of Service·Privacy Policy