Regal
Regal · Regal: Enterprise Voice, SMS and Chat AI Agents for Contact Centers That Qualify Leads, Book Appointments, Run Collections and Resolve Support Calls
Regal is an enterprise platform for building and running AI agents that answer and place phone calls, reply by SMS and web chat, and talk to customers in the browser over WebRTC, handling lead qualification, scheduling, collections, bookings and support for contact centers. It suits large consumer-facing companies in insurance, lending, healthcare, education and home services that run at least tens of human agents and want AI and human agents on one platform alongside their existing CRM and CCaaS.
Best for
Enterprise consumer brands with real contact centers (Regal suggests at least 25 agents and 100,000 call minutes a month) in insurance, Medicare, lending, healthcare, education and home services that want AI agents for lead qualification, scheduling, collections and support on the same platform as their human agents, dialer and customer data
Not ideal for
Small businesses that want a self-serve AI receptionist with published per-minute pricing, developers who want to assemble their own voice stack in code, teams that need on-premise deployment, and support organizations whose volume is mostly email and chat tickets inside a helpdesk
Who it's for
Contact center, CX, sales operations and enrollment leaders at large consumer-facing companies in insurance, Medicare, lending and banking, healthcare, education, legal and home services, plus the BPOs that run their calls
Capabilities
- Phone AI agents for inbound lines and outbound calling (triggered calls, batch dialing, outbound routing), with answering machine detection and voicemail drop
- Goal Oriented Agents (launched October 27, 2026): goal nodes that define required outcomes and prerequisites, including plain-language conditions, with a fast actor model holding the conversation and a director model checking progress after every turn and moving the call to the next stage
- No-code agent builder with single-state and multi-state prompts, drafts and versioning, voice and personality settings, a knowledge base and per-agent LLM choice
- In-call actions to transfer to humans or other AI agents, schedule callbacks, gather dates and times, authenticate the contact, press IVR digits, start or stop recording, update the contact profile and book on Cal.com, Google Calendar or Outlook
- Custom actions that call your APIs and JavaScript functions hosted and edited inside Regal, with captured values reusable as variables later in the call
- SMS agents, an embeddable chat widget or SDK, and a WebRTC voice widget that lets website or app visitors talk to the agent with one click and switch between chat and voice without losing context
- Regal Copilot, a chat interface that generates an agent from a description, script or transcript, stress-tests it with an auto-generated test suite and simulations, and suggests fixes from live transcripts that you approve in one click
- Native A/B testing of agent variants with traffic allocation, conversation intelligence, automated QA scoring, live call monitoring and analytics
- Journey Builder and a unified customer profile that trigger calls, texts and webhooks from real-time customer events sent by tools such as Segment, mParticle, Braze, Iterable or Customer.io
- Native Twilio-based telephony with number management, branded caller ID, local presence and spam remediation, or SIP connection to an existing CCaaS such as NICE or Five9, passing context in SIP headers
- Accredited Five9 AI Agent Connect partner: Five9 stays on the call, Regal leaves when done, and Five9 asks Regal's Call Handoff API whether to end the call or route it to a specific agent, skill or queue
- Unified agent desktop for human agents with phone, SMS and email in one inbox, a sales dialer, and an embeddable iFrame for Salesforce or other web apps
- Choice of more than 20 LLMs from OpenAI, Google and Anthropic plus GLM-4.7 in beta, a dozen of them benchmarked in the docs, with GPT-5.6 Luna as the default, and around 30 voice languages and locales plus a beta multilingual mode that detects 10 languages
- Regal MCP server for Cursor, Claude, ChatGPT and custom clients, with Okta-based sign-in, brand-scoped tools to create and edit agents and variants, run tests and read simulations, and an explicit confirmation required to publish an agent live
- REST API for contacts, events, messages, routing and call handoff, plus real-time reporting webhooks
Limitations
- No public pricing: the pricing page asks for a form to build a custom quote, and price depends on the features used and contract length
- Built for enterprises: Regal's own pricing FAQ gives at least 25 agents and 100,000 call minutes a month as a good proxy for fit
- No self-serve sign-up or free trial is documented; access starts with a demo and a sales-led implementation
- Goal Oriented Agents were launched to existing customers through their Regal CSM rather than as a self-serve feature
- Model choice changes cost: the docs say the default GPT-5.6 Luna is included in the base voice agent minute price, while GPT-5.6 Terra costs about 10 times as much per turn
- SIP telephony runs through Twilio, and SIP transfers between Twilio accounts that are not owned by the same parent organization are not supported, so cross-account setups use Application Connect or PSTN transfers
- The multilingual voice mode that detects the caller's language is in beta and covers 10 languages; selecting a language does not change a voice's accent, so other languages may need a native voice
- The external MCP server exposes only a subset of the tools in-app Copilot uses, and full ChatGPT MCP support needs a ChatGPT Business or Enterprise workspace
- No self-hosted or on-premise deployment is documented
Use cases
- Calling every new insurance, lending or education lead as soon as it arrives, qualifying it and warm-transferring qualified callers to a licensed agent
- Taking auto insurance claims or roadside assistance calls where callers give many details at once, until the agent has enough to open a case
- Running Medicare enrollment or HIPAA-covered medical intake calls with fixed prerequisites the agent must confirm before acting
- Scheduling, confirming and rescheduling appointments by phone and SMS to reduce no-shows
- Working collections and loan servicing calls and texts at scale while keeping humans for exceptions
- Answering inbound support calls with an AI agent in front of an existing Five9 or NICE contact center and routing escalations to the right queue with context
Our take
Regal's strongest argument is that it replaces more than the conversation. A contact center that buys it gets voice, SMS, chat and WebRTC agents plus the telephony, caller ID, dialer, journey orchestration, A/B testing and human agent desktop around them, which matters most for outbound-heavy, regulated businesses such as insurance, Medicare and lending where a lead has to be reached fast, qualified correctly and handed to a licensed person. Goal Oriented Agents are a sensible answer to the usual choice between loose single-prompt agents and rigid flows, though they are new and gated through customer success. The documentation is unusually frank about model tradeoffs and Twilio-based SIP constraints. The cost is the enterprise model: no public pricing, a stated fit of 25 or more agents and 100,000 or more minutes a month, and a forward-deployed implementation, so smaller teams and developers who want to build their own voice stack should look elsewhere.
Who should use it
Contact center and CX leaders at large consumer brands, especially in insurance, Medicare, lending and banking, healthcare, education and home services, that handle high call volumes, need SOC 2, HIPAA and PCI coverage, and want AI agents to qualify, schedule, collect and support while working alongside their human agents, dialer and existing CCaaS.
Who should skip it
Small businesses and solo operators, teams that need published prices or a self-serve trial, developers who prefer to build voice agents in code on their own speech and telephony providers, organizations that require on-premise deployment, and support teams whose work is mostly helpdesk tickets rather than calls.
Strengths
- AI agents, telephony, dialer, journey orchestration, A/B testing, QA and a human agent desktop on one platform, so AI and human agents share the same customer data
- Goal Oriented Agents add a director model that tracks required outcomes on every turn, aimed at long intake and qualification calls that rigid flows handle poorly
- Works in front of an existing CCaaS over SIP, with an accredited Five9 AI Agent Connect integration and Salesforce embedding
- Official SOC 2 Type II, HIPAA, GDPR, CCPA and PCI DSS compliance, SSO and role permissions, and a 99.9% voice uptime SLA
- Wide model choice with published benchmark scores and evaluation notes for a dozen models, plus a statement that LLM partners may not train on customer data
- Implementation handled by forward deployed engineers, plus an MCP server and API for teams that want to manage agents from their own tools
Weaknesses
- Custom, sales-led pricing with no public rates, free tier or self-serve trial
- Aimed at contact centers of roughly 25 or more agents and 100,000 or more call minutes a month, which rules out most small businesses
- The newest agent design, Goal Oriented Agents, is gated through customer success at launch
- More capable LLMs can cost several times more per turn than the default model
- Broad platform with many modules to configure, from telephony registration to journeys and dialers
Regal pricing
Custom
Custom quote
- Dedicated implementation team and emergency support
- SOC 2 Type II, HIPAA, GDPR, CCPA and PCI DSS compliance
- 99.9% uptime SLA on voice infrastructure
- Conversation analytics, call scoring and observability
- Automatic access to new LLMs, voices and features
- Native CRM, dialer and data connectors, with integrations handled by forward deployed engineers
Note: Regal does not publish prices; you request a custom quote through a form. Its pricing FAQ says you pay for the features you use, with discounts for higher committed spend and longer contracts, and that a good fit is at least 25 agents and 100,000 call minutes a month. The docs say the default model, GPT-5.6 Luna, is included in the base AI voice agent minute price, while some other models cost more per turn. No free tier or self-serve trial is documented.
Technical specs
Modalities
Voice, Text
API pricing
No separate API pricing is published; API and MCP access come with a Regal account under a custom contract
Available models
Where Regal excels
Speed-to-lead for insurance or lending
A Journey Builder trigger calls each new lead as soon as it arrives, the agent qualifies it against your criteria, and qualified callers are warm-transferred to a licensed agent through Regal or an existing Five9 or NICE queue with the collected details attached.
Claims and intake calls
A Goal Oriented Agent lets callers explain an accident or medical situation in their own order, checks every answer against the goals and prerequisites for the stage, and asks only for what is still missing before opening a case or booking the next step.
Appointment scheduling and reminders
Voice and SMS agents book, confirm and reschedule on Cal.com, Google Calendar or Outlook, with A/B tests on timing and wording to reduce no-shows.
AI in front of an existing CCaaS
With Five9 AI Agent Connect or a SIP trunk, Regal's agent answers first, and when it leaves the call the contact center platform asks Regal where to route it next and what to show the human agent.
Regal vs. competitors
Regal vs. Sierra
Sierra is an enterprise platform for branded customer-facing agents that take actions across support, sales and account workflows on voice, chat, email and messaging. Regal is centered on the contact center itself, bundling AI agents with telephony, branded caller ID, dialers, event-driven journeys and a human agent desktop, and leans further into outbound lead qualification, enrollment and collections.
Regal vs. Decagon
Decagon builds enterprise AI support agents across chat, voice, email and SMS, defined through natural-language Agent Operating Procedures and focused on resolving support conversations. Regal covers support too, but is voice-first and built for phone operations, including outbound calling, number management and spam remediation, dialers and journeys for sales, scheduling and collections.
Regal vs. Harmony
Harmony is monday.com's voice AI for revenue teams, with an AI dialer, conversational IVR and CRM write-back for outbound and inbound sales calls. Regal overlaps on lead qualification and dialing but targets larger regulated contact centers with custom pricing, adds SMS, chat and WebRTC agents, journeys and a human agent desktop, and also covers support, scheduling and collections.
Regal vs. Vapi
Vapi is an API-first platform for developers who assemble and run voice agents in code with their own choice of providers. Regal is a managed, no-code enterprise platform where CX teams build agents with Copilot and forward deployed engineers and get telephony, compliance and contact center tooling included.
Frequently asked questions
What is Regal?
Regal is an enterprise platform for building, deploying and managing AI agents for contact centers. Its agents handle phone calls, SMS, web chat and in-browser WebRTC voice for lead qualification, appointment setting, collections, bookings and customer support, and the platform also includes telephony, a journey builder, a unified customer profile, A/B testing, QA and a desktop and dialer for human agents.
How much does Regal cost?
Regal does not publish prices. Its pricing page asks you to fill out a form for a custom quote, says you pay for the features you use, and offers discounts for higher committed spend and longer contracts. All plans include a dedicated implementation team, SOC 2 Type II, HIPAA, GDPR, CCPA and PCI DSS compliance, and a 99.9% uptime SLA on voice infrastructure.
What are Regal's Goal Oriented Agents?
Launched on October 27, 2026, Goal Oriented Agents are voice agents you point at an outcome instead of a script. You define goal nodes with the goals to reach (collected fields or tools such as scheduling a callback) and prerequisites (data fields or plain-language conditions). On every call a small, fast actor model replies to the caller while a director model checks after each turn which goals are met, then guides the actor or moves the call to the next stage. Existing customers request access through their Regal CSM.
Does Regal work with my existing contact center software?
Yes. Regal has its own Twilio-based telephony, but it can also sit behind another provider such as NICE or Five9 over a SIP trunk, receiving calls and transferring them back with context in SIP headers. Regal is an accredited Five9 AI Agent Connect partner, and its agent desktop can be embedded in Salesforce. Its integrations list also includes Talkdesk and 8x8.
Is Regal HIPAA compliant?
Regal's pricing page lists SOC 2 Type II, HIPAA, GDPR, CCPA and PCI DSS compliance in all plans, and its enterprise page adds SSO and role-level permissions. Regal also says it does not train proprietary models and that its agreements with LLM providers such as OpenAI, Anthropic and Google prohibit training on customer data.
Which channels and languages does Regal support?
Regal agents work on inbound and outbound phone calls, SMS, an embeddable chat widget and a WebRTC voice widget for websites and apps. Voice agents can be set to around 30 languages and locales, including several English and Spanish variants, and a beta multilingual mode detects and answers in 10 languages. SMS agents reply in whatever language the prompt allows.
Which AI models does Regal use?
Regal lets you pick the LLM per agent. Its docs benchmark a dozen models on task completion, speed and relative cost, including OpenAI GPT-5.6 Luna (the default), GPT-5.6 Terra, GPT-4.1, Google Gemini 3.6 Flash, Anthropic Claude Haiku 4.5 and Claude Sonnet 4.6, and GLM-4.7 in beta, and list further models that are available but not yet benchmarked, such as other GPT-5 versions, Claude Sonnet 4.5 and the GPT Realtime models.
Can I manage Regal agents from Claude, ChatGPT or Cursor?
Yes. The Regal MCP server connects MCP clients such as Cursor, Claude and ChatGPT to your Regal account through the same Okta login. It can list and inspect agents, create agents and A/B variants, set traffic splits, create tests and read simulation transcripts; publishing an agent live requires an explicit confirmation.
Regal vs Sierra: what is the difference?
Both are sales-led enterprise platforms for customer-facing AI agents. Sierra is positioned as a branded agent layer across support, sales and account workflows on many channels. Regal is built around the phone-heavy contact center: outbound and inbound calling, telephony and caller ID, dialers, journeys triggered by customer events and a desktop for human agents, with use cases weighted toward lead qualification, enrollment and collections as well as support.
Integrations & fit
Alternatives to consider
Sierra
Enterprises that want branded, action-taking AI agents as the front door to their customer experience, with deep control over tone and guardrails
CustomDecagon
Large consumer brands automating high-volume, multi-channel customer support with on-brand, measurable AI agents
Custom
monday.com
Mid-market and enterprise sales, insurance, finance and real estate teams in the US, UK, Australia or Singapore that call large lead volumes and want an AI SDR, dialer and CRM write-back in one vendor-managed product with SOC 2, ISO 27001 and HIPAA BAA options
PaidAbout Regal
Regal started as an outbound phone and SMS sales platform for consumer brands and has turned into what it now calls an enterprise voice AI agent platform for contact centers. Teams build AI agents in a no-code builder, either as a single prompt, as a multi-state flow, or since October 27, 2026 as a Goal Oriented Agent: you define goal nodes that list what the call must achieve (collected fields or tools such as scheduling a callback) and the prerequisites before acting, and two models share each call. A small, fast actor model replies to the caller while a director model checks after every turn which goals are met, so a caller who gives several details at once is not asked for them again; existing customers get access through their Regal CSM. Agents act during the call through built-in actions (transfer, schedule a callback, book on Cal.com, Google or Outlook calendars, authenticate the contact, press IVR digits, drop a voicemail) and custom API actions or JavaScript functions, and the same agent logic can run on phone, SMS, an embeddable chat widget and a one-click WebRTC voice widget. Around the agents sits the rest of a contact-center stack: native Twilio-based telephony with branded caller ID and spam remediation, a Journey Builder that triggers calls and texts from customer events, a unified customer profile fed by tools like Segment, Braze or Salesforce, a sales dialer and a unified desktop where human agents take transfers, plus A/B testing of agent variants, automated QA and Regal Copilot, which drafts agents from a description or transcript, generates test suites and proposes fixes from live transcripts. Companies that keep NICE, Five9 or another CCaaS can route calls to Regal over SIP, and Regal is an accredited Five9 AI Agent Connect partner. The tradeoffs are the enterprise model: pricing is a custom quote, Regal says a good fit is at least 25 agents and 100,000 call minutes a month, and implementation runs with Regal's forward deployed engineers rather than self-serve sign-up.
Updates from Regal
Regal added Goal Oriented Agents, built from goal nodes that list the outcomes and prerequisites for each stage of a call. A fast actor model replies to the caller while a director model checks progress after every turn and moves the call forward, aimed at complex intake and qualification such as insurance claims, Medicare enrollment and mortgage qualification. Existing customers get access through their Regal CSM.
Are you the founder? Claim this listing →