You're building a SignalWire app. You need to figure out how to structure a SignalWire Markup Language (SWML) call flow, implement context-aware transfers, or set up SignalWire AI Gateway (SWAIG) functions. You could dig through documentation, search GitHub examples, or ask in Discord.

Or you could just ask Claude.

The new SignalWire skill for Claude Code gives you an AI pair programmer that knows SignalWire inside and out. It understands the REST APIs, writes SWML, explains AI agent patterns, and helps you avoid common mistakes.

What is the SignalWire Skill for Claude Code?

Think of it as having a SignalWire expert sitting next to you. Need to implement appointment reminders? It'll write the complete flow. Building an AI agent for customer service? It knows the patterns. Stuck on Subscriber configuration? It explains the concepts and shows you the code.

The skill covers everything you need to build production SignalWire apps:

Voice AI and SWML

• Write complete call flows in SWML

• Structure AI agent prompts that actually work

• Implement SWAIG functions for external integrations

• Handle context preservation across transfers

REST APIs and SDKs

• Make outbound calls with proper authentication

• Send SMS with Campaign Registry compliance

• Manage video rooms and tokens

• Work with the Relay SDK for real-time communication

Production Patterns

• Secure sensitive data with metadata patterns

• Implement loop protection in call flows

• Set up webhook testing workflows

• Add post-prompt analytics

How it works

Install the skill and Claude automatically knows when you're working on SignalWire code. Ask it to generate SWML, explain a concept, or review your implementation.

Once installed, just start coding. Claude recognizes SignalWire context and provides relevant help.

Example prompts that work:

• "Create an IVR that routes to sales, support, or billing"

• "Write a SWAIG function that checks inventory and returns availability"

• "Show me how to implement context-aware transfers"

• "Review this AI agent prompt and suggest improvements"

• “Make me an AI agent that can call and order me a pizza.”

What makes the SignalWire Skill different

This isn't generic AI help. The skill understands SignalWire's specific patterns and best practices.

It knows to use metadata for sensitive data instead of putting it in AI prompts. It writes SWML with proper loop protection. It implements SWAIG functions that build knowledge progressively. It avoids the over-prompting patterns that make AI agents brittle.

When you ask for a restaurant ordering system, you get working code that follows production patterns—metadata for payment info, progressive knowledge building through SWAIG, context preservation for transfers.

Real examples

Building an AI agent: You ask Claude to create a support agent. It generates a complete implementation with:

• Natural, conversational prompts (no programming constructs)

• SWAIG functions for ticket creation and lookup

• Metadata handling for customer information

• Context-aware transfer patterns

• Post-prompt analytics configuration

Creating a video conference: Request a video room setup. You get:

• Secure token generation (never exposes API credentials to browser)

• Room configuration with proper layouts

• Browser SDK integration code

• WebSocket connection handling

SMS with compliance: Ask about SMS delivery. It explains The Campaign Registry requirements and generates:

• Brand and campaign registration steps

• Message templates with proper opt-out language

• Delivery status tracking

• Rate limiting patterns

Getting started

The skill works immediately after installation. No configuration needed. No separate documentation to read.

Start with simple requests. "Create a basic IVR." "Show me how to authenticate." "Explain SWAIG functions." As you work, ask more specific questions about your actual implementation.

The skill includes all the workflow documentation you need—authentication, calling, messaging, video, AI agents, webhooks. When Claude answers your question, it's pulling from comprehensive, tested patterns.

What you get with the Claude Code Skill

Production-ready code. Not pseudocode. Not "here's the general pattern." Complete, working examples with proper error handling, security patterns, and best practices built in.

The code uses modern SignalWire APIs. You get SWML in YAML or JSON. REST API calls with proper authentication. Relay SDK patterns for real-time communication. AI Agents SDK for voice AI.

Every example follows the same production patterns you'll see in SignalWire's own implementations.

Try it

Install the skill and try these starter prompts:

1. "Create a simple IVR with loop protection"

2. "Write an AI agent that collects customer info before transferring"

3. "Show me how to send SMS with delivery tracking"

4. "Implement MFA using phone calls"

Watch how Claude not only generates the code but explains the patterns and points out potential issues.

The skill is open source. You can review the documentation, suggest improvements, or adapt it for your team's specific patterns.

Download the SignalWire Claude Code Skill

The SignalWire skill is available now at https://github.com/signalwire/signalwire-claude.

Install it once. Use it every time you build with SignalWire.