> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.rydhm.ai/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.rydhm.ai/_mcp/server.

# Concepts

Read this page once. Every other page uses these words the same way.

## Workspace

A **workspace** is your team's space in Rydhm. It holds your experts, their knowledge, tools, phone numbers, conversations and billing. Everything you create belongs to one workspace, and the top bar always shows which one you're in.

## Members and roles

People join a workspace as an **admin** or a **member** (a workspace can also have an **owner**).

* **Admins** create, configure, make live and delete experts, manage knowledge and tools, and see billing.
* **Members** talk to experts, read their settings and conversations, and can edit an expert's name, description, picture and background. Tabs an admin manages show a lock and read-only values.

See [Members and roles](/workspace/members).

## Expert

An **expert** is one AI agent. It has:

| Part                                   | What it decides                                        | Where you set it                                 |
| -------------------------------------- | ------------------------------------------------------ | ------------------------------------------------ |
| **Persona** and **first message**      | Who it is, how it behaves, what it says first          | [Profile tab](/experts/persona)                  |
| **Variables**                          | Blanks like `{{name}}` filled in for each conversation | [Profile tab](/experts/persona#variables)        |
| **Guardrails**                         | How strictly it stays on topic, rules, blocked topics  | [Profile tab](/experts/guardrails)               |
| **Knowledge**                          | What it knows                                          | [Knowledge tab](/knowledge/overview)             |
| **Voice**, **languages** and **style** | How it hears, thinks and speaks                        | [Voice tab](/experts/voice-engines)              |
| **Call experience**                    | Turn-taking, silence, call length, background sound    | [Voice tab](/experts/guardrails#call-experience) |
| **Tools** and **specialists**          | What it can do, and who it can hand a call to          | [Tools tab](/experts/tools)                      |

The expert editor also has **Phone**, **Meetings**, **Share** and **Analysis** tabs for its channels and results. An expert has an **Active** switch on its Profile tab: when it's off, customers can't talk to it.

## Voice style

The **style** is how an expert talks, picked on its **Voice** tab. There are three, and each runs one of Rydhm's voice engines:

* **Balanced** (the **Standard** engine): a transcriber, a language model and a voice, each chosen separately. The default, and it works everywhere, including meetings.
* **Fastest** (**Realtime (Google)**, Gemini Live): one Google model hears and speaks, with the shortest pause.
* **Most natural** (**Realtime (OpenAI)**, GPT-Live): OpenAI's full-duplex model, which listens while it speaks. One-on-one conversations only.

Under **Advanced** you can pick the engine and its models yourself. See [Voice engines](/experts/voice-engines).

## Knowledge

**Knowledge** is the material an expert answers from: files, pasted text, pages imported from a website, and question-and-answer sets. An expert has its **own** knowledge, and can also use **shared** knowledge that several experts read at once. When an expert's knowledge is small, it reads all of it on every turn; when it is large, it searches it for each question. See [Knowledge overview](/knowledge/overview).

## Tools

A **tool** is something an expert can do during a conversation: end the call, look up a calendar, send an email, search the web, or call your own API. Some tools ask the caller to confirm before they act. See [Tools](/experts/tools).

## Channels

A **channel** is a place people reach an expert:

* inside the app: the Test panel, chat and calls from the **Experts** page;
* a [share link](/channels/public-links) and the [website widget](/channels/web-widget);
* [phone numbers](/channels/phone) and outbound [campaigns](/channels/campaigns);
* [meetings](/channels/meetings) on Google Meet, Zoom and Teams;
* the [API, SDKs, MCP and agent-to-agent access](/developers/overview).

## Visibility

Every expert is a **Draft**, **Live** or **Live and listed**:

* **Draft**: only your team can talk to it. Every new expert starts here.
* **Live**: anyone with its link can talk to it, by voice or chat, and your website widget answers.
* **Live and listed**: anyone can, and it is listed in the public directory.

You change this with **Make live** in the expert's header (it reads **Live** once the expert is live). In the API the same three are the `visibility` values `private`, `unlisted` and `public`. Usage on a live expert is billed to your workspace.

## Members and visitors

Rydhm treats two kinds of caller differently:

* **Members** are signed in to your workspace: the Test panel, in-app chat and calls, and meetings an expert joins for you.
* **Visitors** are everyone else: share links, the website widget and phone callers.

A visitor's conversation can use the expert's knowledge, its conversation controls and web search. Actions on your workspace (calendar, email, your own API tools, outside tool servers, hand-offs) run only for members. See [Tools](/experts/tools#who-can-use-which-tools).

## Conversation

A **conversation** is one chat, call or meeting with an expert. Rydhm keeps its transcript, and can summarize and analyze it. All of them are listed under **Conversations** in the sidebar. See [Conversations](/quality/conversations).

## Version

Each time an expert is created, saved, made live or restored, Rydhm keeps a **version**: a snapshot of its settings, with who made it and what changed. You can compare versions and restore one. See [Versions and testing](/experts/versions).