> For the complete documentation index, see [llms.txt](https://docs.praiz.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.praiz.io/query-and-analyze/mcp-server.md).

# MCP server

Connect your conversational database to Claude, ChatGPT or any MCP-compatible tool, and query all of your calls in plain language.

You query your calls from your AI tool, without going through the Praiz interface.

{% hint style="info" %}
Before you start: a Praiz account with transcribed calls, and an account on an MCP-compatible AI tool.
{% endhint %}

## What MCP is

MCP (Model Context Protocol) is a standard that lets an AI tool reach an external data source. The Praiz MCP server exposes your conversations: calls, transcripts, participants and agent outputs. The AI tool answers from the content of your calls, not just from its general knowledge.

## What the connector can do

* **Search and read your calls**: search by keyword, participant or period; read a full transcript or an excerpt around a given moment; participants and comments of a call.
* **Use your agents' data**: list your agents and their fields, values extracted on a call, bulk export of an agent's values with filters.
* **Set up agents** (admins): create an agent, add or edit its fields, choose their position and reorder them, and set or remove each field's CRM sync, directly from your AI tool.
* **Read your connected CRM**: your CRM's objects and properties, with their type and possible values, so you can map a field without opening your CRM.
* **Explore your workspace**: users, teams, playlists, tags, webhooks (admins) and usage statistics.

Ready-to-copy examples are on the [What to ask the MCP connector](/query-and-analyze/mcp-server/what-to-ask-the-mcp-connector.md) page.

## Connect the MCP server

The server address is the same for everyone: `https://mcp.praiz.io/mcp`. There is no key or token to generate. You authenticate with your Praiz account when you connect. The address is also shown in Praiz, under "Settings", "Integrations" tab.

![Integrations tab in Settings, with the MCP server address](https://3785996597-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FU54oKzNb3EyinriyUP5P%2Fuploads%2FO6P7oIzwSQFa9zMh0XOa%2F05-settings-integrations.png?alt=media)

{% tabs %}
{% tab title="Claude" %}

1. In Claude, open the settings then "Connectors" (claude.ai/customize/connectors).
2. Click "Add custom connector".
3. Paste `https://mcp.praiz.io/mcp` and confirm.
4. Sign in with your Praiz account in the window that opens and choose your workspace.

If adding connectors is blocked in your organization, ask your Claude admin to allow it.
{% endtab %}

{% tab title="ChatGPT" %}

1. In ChatGPT, open the settings then "Connectors".
2. Turn on developer mode ("Developer mode"), which is required for custom MCP connectors.
3. Create a connector with the address `https://mcp.praiz.io/mcp`.
4. Sign in with your Praiz account.
   {% endtab %}

{% tab title="Cursor" %}

1. Open "Settings", then "Tools & MCP", then "New MCP Server".
2. Declare a server with the URL `https://mcp.praiz.io/mcp`.
3. Click "Needs login" and sign in with your Praiz account.
   {% endtab %}

{% tab title="Other tools" %}
Add a remote MCP server with the URL `https://mcp.praiz.io/mcp` and authenticate with your Praiz account. The server works with n8n (MCP Client node) and any MCP client that supports sign-in authentication. Codex CLI: declare the server with this address, then run `codex mcp login praiz` (praiz being the name you gave the server) to sign in with your Praiz account. For Gemini and Microsoft Copilot the setup is longer: contact support.
{% endtab %}
{% endtabs %}

What to expect: Praiz appears in the list of connected tools. Your AI tool can search your calls.

## What you can ask

* Which objections come up most often on our lost deals?
* What do our prospects say about this competitor?
* What product feedback came up this quarter?
* Prepare a summary of our exchanges with this customer before my meeting.
* Export the transcripts of the last 30 days.

More examples, sorted by use case: [What to ask the MCP connector](/query-and-analyze/mcp-server/what-to-ask-the-mcp-connector.md).

You can combine the Praiz MCP server with other MCP servers in the same tool, for example your CRM's. That lets you cross conversations with CRM data.

## Scope

The MCP server gives access to all of your conversations, within your Praiz permissions and the workspace you chose when connecting. Reading never changes your data; only admins can create or edit agents. Lists are paginated at 100 items at most.

| You want to                                 | Use                                          |
| ------------------------------------------- | -------------------------------------------- |
| Dig into one point on one specific call     | [Ask Praiz](/query-and-analyze/ask-praiz.md) |
| Surface trends, inside Praiz                | [Insights](/query-and-analyze/insights.md)   |
| Query your whole database from your AI tool | MCP server                                   |

## Troubleshooting

* The AI tool asks for authorization again: the session expired. Sign in again.
* Results are empty: check that you are connected to the right Praiz workspace. Disconnect and reconnect the connector if needed.
* A call is missing: check that it is visible to you in Praiz. The MCP server respects [private mode](/administration/private-mode.md).
* A recent tool is missing: refresh the connector's tool list in your AI client.

## Webhooks

To receive agent outputs, the transcript and a call's metadata in your own system automatically, use [webhooks](/integrations/webhooks.md). An admin sets them up under "Settings". Praiz does not expose a public API.