Your voice. Your agent. Your workspace.

Build voice agents with Pipecat and Inworld. Configure instructions, publish revisions and connect approved tools in an invited workspace.

What is Conversational AI?

From a script to a conversation

Give visitors a voice interface to your business. Define what your agent should say, what it can access and when your team should review the conversation.

Before Conversational AI

"Press 1 for billing, press 2 for ...". Rigid IVR phone trees that frustrate customers before they reach a human.

Rule-based chatbots that follow scripts. One unexpected question and the whole experience falls apart.

Customers repeat themselves every time they switch channels, phone to chat to email. Nothing carries over.

Scaling means hiring. More agents, more shifts, more training, more cost. Quality varies person to person and call to call.

Single-language support, or clunky handoffs between language-specific teams that slow resolution and lose key context.

With our Conversational AI platform

Set instructions and a greeting for your agent.

Attach approved knowledge and tools to a published revision.

Use the configured website widget for browser conversations.

Review conversation history and refine the next revision.

Choose a voice and language available in your workspace.

Build, test, then publish

Open workspace

Start in your invited workspace. Configure an agent, review its access and publish a revision when it is ready for an enabled call environment.

01

Define its instructions

Write a greeting and instructions. Attach approved knowledge for the questions your agent should answer.

02

Review its tools

Attach approved private tools or MCP connectors. Workspace policies determine which actions are available.

03

Review and improve

Review call history and publish a new revision after updating the agent.

Conversational AI use cases

  • Answer common questions using your approved instructions and knowledge.

Support guideCustomer support
Open workspace

Interface example

Shape how your agent speaks

Choose an available voice and write instructions that reflect your brand. Listen to test conversations before publishing changes.

01

Your instructions

Define tone, scope and how the agent should respond to unfamiliar requests.

02

Approved actions

Publish only the tools approved for your workspace.

03

Conversation review

Use recorded conversation history to identify where instructions need improvement.

04

Controlled access

Use member roles and workspace policies to control editing and execution.

Voice conversations

Pipecat coordinates the voice conversation with the configured speech services.

Workspace voices

Select from voices available to your workspace.

Language settings

Choose a supported language and test it with your selected voice.

A workspace your team controls

Invite members and assign owner, admin, editor or viewer roles.

Getting started with conversational AI

Create on the platform

Sign in with an invitation to configure agents in your team workspace. Calling requires workspace enablement.

Sign up

Website integration

Use the configured embed and approved connectors. Review setup documentation for your environment.

Frequently asked questions

AI Communication Platform