LiveKit

LiveKit is an open source framework and developer platform for building, testing, deploying, scaling, and observing voice, video, and physical AI agents in production. A voice agent is defined in code, in Python or Node.js, by choosing the speech to text, language model, and text to speech services for a session, with automatic turn detection and interruption handling built in, so a caller can speak naturally and break in while the agent is talking. Quickstarts and starter apps get a simple voice agent running through a web or mobile app, and a telephony integration lets the same agent make or take phone calls. The stack is deliberately complete. It combines the open source Agents framework for building and testing, an inference gateway for reaching speech, language, and voice models, the LiveKit Cloud platform for deploying and scaling agents, phone numbers and SIP integrations for telephony, and full-stack observability for every agent session. The flow is simple to reason about: the user speaks through an app, a browser, or a phone call, the speech streams to the agent, the agent runs your custom business logic, and it responds back to the user. Because the logic is your code, teams can fine-tune agent behavior for each workflow and use any model provider, a level of control the vendor's healthcare and sales agent customers cite on its homepage. Scale and trust claims are stated plainly by the vendor. It lists GDPR, SOC 2 Type 2, and HIPAA among its enterprise credentials, and publishes figures it attributes to its own platform, including hundreds of thousands of developers, billions of calls a year, and regions around the world, and it notes that OpenAI built the Advanced Voice mode of ChatGPT on LiveKit Cloud. It also claims the platform can run millions of concurrent calls. Those are vendor figures, presented as the vendor presents them, and a buyer should test latency and concurrency against their own call pattern before committing. Pricing starts with a free account that needs no credit card and includes a monthly allotment of agent session minutes, which the homepage states in full. Paid plans scale the allotments for agent session minutes, inbound calling minutes on US local numbers, and inference credits for calling popular models, and a pricing calculator on the pricing page previews the per-minute cost of running an agent on LiveKit Cloud, with separate detailed pricing for language, speech to text, and text to speech models through LiveKit Inference. Teams that outgrow the standard plans can ask about custom pricing. LiveKit fits developers building AI phone agents who want an open source core, their choice of models, and managed cloud deployment and telephony rather than a closed, prebuilt receptionist.

AI Phone Agents

Automation Customer Support Speech

Visit Website

Similar listings in category

Jobber

Field-service management platform with Jobber AI Receptionist, online booking, client management, scheduling, quoting, invoicing, and marketing tools.

Business Tools Service Business AI AI CRM & Lead Management AI Chatbots & Website Receptionists Local-Service AI Tools Missed-Call Text-Back & Follow-Up Appointment Booking & Scheduling AI Phone Agents AI Tools for Contractors and Home Services Service Business

Retell AI

Retell AI is an AI voice agent platform for contact centers, in production at more than 4,000 businesses according to its homepage. Its agents answer inbound calls and place outbound calls, qualify leads, book appointments, check order status, and route or transfer callers to a live agent when needed. Retell is built for real contact center demands such as transfers, phone trees, outbound volume, and back end systems that have to work on every call. Features include agentic warm transfers, where the agent briefs the receiving person before connecting the customer, IVR navigation through complex phone trees and extensions, and a mix of deterministic and generative responses so critical steps such as disclosures and verification follow exact requirements while the rest of the conversation stays natural. During a live call, Retell agents can execute custom logic, access real time data, and complete tasks across business systems, from booking appointments to updating records. Persistent memory keeps a running brief on every contact across calls and texts, so returning callers can pick up where they left off. The platform also offers burst concurrency, sentiment analysis, AI quality assurance, live monitoring, and a shared caller history that can sync with an existing CRM or run in Retell CRM. CX teams build agents in a no code studio while developers extend them with APIs and SDKs, and every call is transcribed for review. Retell cites response times around 600 milliseconds and compliance with HIPAA, SOC 2 Type II, GDPR, and ISO 27001, with SSO, personal information redaction, and role based access control. On its pricing page, Retell describes true pay as you go pricing for voice agents at $0.07 to $0.31 per minute depending on the models, voice, and add ons chosen, with enterprise pricing custom, no platform fee, and 20 concurrent calls included. New accounts receive $10 in free credits on signup according to the homepage.

AI Phone Agents Automation Customer Support Speech

Vapi

Vapi is a platform for building advanced voice AI agents, aimed at builders who want to configure agents on their own terms. Its homepage tagline is voice agents for builders, with the promise that teams can build, deploy, and improve voice agents as fast as they ship. Vapi presents itself as one unified platform for all of a team's voice agents, combining orchestration, real time monitoring, and enterprise grade configurability built for scale. Builders configure everything from voice and conversation flow to telephony and integrations, while Vapi handles the infrastructure so a team can move from prompt to production fast, with build, test, and deploy described as taking minutes. Once agents are live, Vapi is built around continuous improvement. Teams can track what is working across every call, surface what is not, and keep improving the experiences that drive their business forward. The homepage reports scale figures of 1 billion calls supported, 2.5 million or more agents launched, and 1 million or more developers, with average latency under 500 milliseconds. A customer quote from Ring on the homepage says that customer went from zero to production in two weeks, that 100 percent of its inbound volume now runs through Vapi, and that customer satisfaction scores have improved while its high bar for support was maintained. For enterprises, Vapi lists a support SLA with contractual uptime and performance guarantees and reserved capacity sized to volume, dedicated deployment support with a dedicated engineer embedded in the team to get live in a week, and scalable infrastructure that can scale to millions of calls. Security and control features listed include SSO, OAuth, and role based access control, AI guardrails built in to prevent model hallucinations and ensure data integrity, and compliance described as SOC 2, HIPAA, and PCI compliant, with 99.9 percent uptime stated for enterprise clients.