SignalWire, the leader in Programmable Unified Communications (PUC), today announced the release of its Agents SDK, a Python-native toolkit for building, deploying, and scaling ultra-low latency AI voice agents. With built-in support for SIP, PSTN, WebRTC, and messaging, developers can go from prototype to production without telecom complexity or stitching together multiple APIs.
“We built FreeSWITCH to give developers control over telecom infrastructure,” said Anthony Minessale, Founder and CEO of SignalWire. “With Agents SDK, we’re giving them a new kind of control: one that lets them focus entirely on building intelligent voice experiences without worrying about SIP complexity, API orchestration, or infrastructure scaling.”
Key features of the Agents SDK include:
- Unified AI runtime: Handle prompts, memory, tools, and workflows in a single service.
- Built-in SIP, PSTN, WebRTC, and SMS: No carrier set-up required.
- Modular Skills system: Add capabilities like web search, wikipedia lookup, or RAG queries with a single line of code.
- Flexible deployment: Run agents as microservices, Lambdas, or CGI scripts – no rewrites required.
- Production-grade performance: ~500ms global latency with built-in failover and routing.
- Replaces SIP middleware: Native SIP support handles transfers, PBX mappings, and call flows without extra layers.
- Interoperability with MCP tools: Use tools built for Model Context Protocol (MCP), an emerging standard for defining AI tool interfaces, without rewriting them.
The SDK deepens integration with SWAIG, SignalWire’s real-time tool gateway similar to MCP, by embedding it directly into agent runtime. A new MCP-SWAIG Gateway bridges the two protocols, allowing SignalWire AI agents to call MCP-based tools without modification. Developers can define tools and backends in one step using a built-in templating engine, while prefab agent archetypes with base personas and logic help teams move faster from idea to deployment.
Until now, AI voice developers were forced to choose between raw media SDKs like LiveKit – which provides audio transport but leaves out telephony, signaling, and runtime logic – or legacy CPaaS platforms that require stitching together GPT calls, SIP gateways, IVRs, and cloud functions. SignalWire’s Agents SDK removes this tradeoff by delivering a full-stack Python runtime with integrated voice, messaging, AI orchestration, and production-ready infrastructure.
“We’ve removed the barriers between AI, real-time voice, and production infrastructure,” Minessale added. “You don’t need a SIP expert or a cloud architect to deploy a voice agent anymore, just a Python developer.”
