> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rippit.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Rippit Agentic Chat + MCP Client

> Connect external MCP servers so Rippit's in-app agent can use their tools alongside your conversation data.

The [Rippit MCP server](/mcp-server/overview) points one way: your AI assistant connects to Rippit and reads your conversation data.

Agentic Chat is the other direction. Rippit's in-app agent is itself an MCP **client**, so you can connect your own MCP servers to it. Once connected, the agent can reach Slack, Intercom, Notion, Salesforce, Snowflake, Databricks, or any MCP server you point it at, in the same conversation where it is analyzing your tickets.

That combination is the point. Rippit already knows what your customers said. Connecting a server tells it where to look for the surrounding context, and where to put the answer.

<CardGroup cols={2}>
  <Card title="Connect a server" icon="plug" href="/agentic-chat/connect-a-server">
    The setup flow, identity choices, and how tool selection works.
  </Card>

  <Card title="Slack" icon="slack" href="/agentic-chat/connectors/slack">
    Search and read channels; deliver reports and alerts.
  </Card>

  <Card title="Intercom" icon="comments" href="/agentic-chat/connectors/intercom">
    Read conversations, contacts, companies, and Help Center articles.
  </Card>

  <Card title="Notion" icon="book" href="/agentic-chat/connectors/notion">
    Search the workspace; read and write pages and databases.
  </Card>

  <Card title="Salesforce" icon="cloud" href="/agentic-chat/connectors/salesforce">
    Query accounts, cases, and custom objects with SOQL.
  </Card>

  <Card title="Snowflake" icon="snowflake" href="/agentic-chat/connectors/snowflake">
    Run governed SQL against your warehouse.
  </Card>

  <Card title="Databricks" icon="database" href="/agentic-chat/connectors/databricks">
    Query Unity Catalog through Databricks SQL.
  </Card>

  <Card title="Custom server" icon="code" href="/agentic-chat/connectors/custom-server">
    Point at any MCP server URL, including your own internal ones.
  </Card>
</CardGroup>

## What this changes

An agent that can only read Rippit answers questions about conversations. An agent that can also reach your other systems answers questions that span them, and then does something with the answer.

* **Bring in the context that isn't in the ticket.** Rippit says 300 customers hit a checkout failure. Salesforce or Snowflake says which of them are enterprise accounts and what they are worth.
* **Ground the analysis in what you have already written.** Match rising contact drivers against your Notion or Help Center coverage to find the gap.
* **Deliver the result where people are.** Post the weekly escalation summary into a Slack channel instead of leaving it in a tab nobody opens.
* **Run it unattended.** Turn work you proved out in chat into an **Agent App** that runs hourly, daily, or weekly with nobody watching. Agent Apps use the same connections as chat, and every run is recorded.

## Setting one up

Connections live under **Connections → MCP Servers** in Rippit, and creating one takes a Full Admin. A connection is made once and is then available to agent runs; individual users do not each authorize it.

Two things are worth knowing before you start, both covered in full on [Connect a server](/agentic-chat/connect-a-server):

* **Connections start read-only.** Only the tools a server declares read-only are enabled. Turning on anything that writes is deliberate and per tool.
* **Scope decides who can use it.** A connection can be yours alone or shared with your organization, and only shared connections are available to Agent Apps.

## What you can connect

Rippit ships tested setup for Slack, Intercom, Notion, Salesforce, Snowflake, Databricks, Google Drive, Guru, Jira and Confluence, Hex, and Replit, and will connect to any other MCP server by URL.

See [Supported servers](/agentic-chat/supported-servers) for the full list, what each one gives the agent, and which ones need setup in their own console first.