SignalWire AI Agent Runtime

What it does

SignalWire AI Agent Runtime takes in an agent definition written in code, such as a Python file that defines steps, scoped tools, and data, plus live phone, SIP, or WebRTC calls. It produces spoken conversational turns, tool-call results, call control actions like transfer or recording, and a structured event stream of each step and decision.

People use it to build voice agents that run on real telephony. The platform owns call state, so the agent server can remain stateless while the model sees only the tools and data scoped to each step. Tool functions run in the developer's own code and return results that update data, move steps, and control the call. Every step is logged to one record that can be inspected for debugging.

It suits developers building production voice agents who want deterministic step control and telephony in one platform. Access is through the server SDKs, which are offered in ten languages, and the Signalwire Space onboarding flow; a free credit is included. Usage is billed at about $0.16 per minute for the runtime plus standard voice transport, with premium TTS adding cost. Limitations include that it is developer-facing rather than a no-code builder, and that voice transport, TTS tier, and carrier fees change the all-in rate.

What it can do

• Define voice agents as Python steps with scoped tools and data per step
• Run AI speech recognition, LLM orchestration, and text-to-speech inside the media path
• Call developer-written tool functions (SWAIG) that return results into the model's context
• Transfer, conference, and record calls as native call control on telephony
• Inspect every step, tool call, and decision on one structured event stream
• Run the same agent logic over PSTN, SIP, and WebRTC

Before you choose

• Voice transport (SIP, PSTN, WebRTC) is billed at standard rates on top of the $0.16/min runtime.
• Premium TTS voices and LLM model choice add per-minute or per-token cost; plan for these in the all-in rate.
• Developer-facing: building agents requires writing code and running a server.

Listed under

• Support
• Voice and phone agents

Get started

Create a SignalWire Space and build a voice agent from the docs and Python example.

1. Create a SignalWire Space at the onboarding page, which includes free credit.
2. Install the SignalWire SDK for your language and define an agent with steps, tools and prompts.
3. Test tool functions locally from the CLI and then deploy the agent to a phone number or SIP endpoint.

Standard STT, LLM processing and TTS are included in the $0.16/min runtime. Voice transport is billed separately at the applicable SIP, PSTN or WebRTC rate, and premium TTS adds per-character charges. Introductory account credit is not an ongoing free tier.

Source: signalwire.com/pricing, checked 30 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 30 Oct 2026. Each source supports part of the overview; availability and prices can change.

• AI Voice agents you can actually put in production - SignalWire ↗signalwire.com
• Voice, Messaging & AI Pricing | SignalWire ↗signalwire.com

Company website

signalwire.com: product, pricing and policy pages

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 30 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.