> ## 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.

# Microsoft Copilot

> Connect Rippit to a Microsoft Copilot Studio agent.

## Add Rippit to an agent

1. Open the **Tools** page for your agent.
2. Select **Add a tool**, then **New tool**.
3. Select **Model Context Protocol**. The MCP onboarding wizard opens.
4. Fill in:

| Field              | Value                           |
| ------------------ | ------------------------------- |
| Server name        | `Rippit`                        |
| Server description | What Rippit does, in a sentence |
| Server URL         | `https://mcp.rippit.com/mcp`    |

Write a real description: Copilot Studio reads it at runtime to decide whether Rippit is relevant to a question, so "Rippit" alone will cause the agent to skip it. Something like *"Analyzes customer support conversation data: volume drivers, themes, sentiment, and QA results"* works well.

5. For authentication, select **OAuth 2.0**.
6. For the OAuth type, select **Dynamic discovery**.
7. Select **Create**, then **Next**.
8. On **Add tool**, select **Create a new connection**, then **Add to agent**.

**Dynamic discovery** is the option to use. Copilot Studio reads Rippit's published metadata and registers itself. You do not need the **Dynamic** or **Manual** options, which exist for servers that can't advertise their own endpoints.

## Signing in

Each user of the agent authenticates with their own Rippit account the first time the agent calls Rippit, via a consent card in the chat. Access is scoped to that user's Rippit permissions.

## Agent instructions

Adding agent instructions is usually required to get a Copilot Studio agent working well with Rippit. Without them, the agent tends to stop and ask the user which column to aggregate or how to filter a table. Those are questions an end user can't answer, and Copilot Studio asks them whenever its orchestrator can't fill a tool input on its own. MCP tools don't have the per-input settings that other Copilot Studio tool types do, so the agent's **Instructions** field is where to fix it.

Paste the block below into your agent's **Instructions** field as a starting point. It works as written, but treat it as a template: edit the role, tone, and output rules to match your team and the questions your users actually ask.

```text wrap theme={null}
# ROLE
Expert Voice of Customer (VoC) analysis partner for Rippit. Customer-written language (what the customer actually said in the conversation) is your ultimate source of truth; AI-generated metadata is supporting evidence only, never primary.

# CONVERSATIONAL STYLE AND OUTPUT PRINCIPLES
Communicate like a natural, insightful, and candid analytical partner, not a rigid form-filler or boilerplate script.
* Answer first and synthesize: Deliver the core answer immediately in clear prose. Blend qualitative context smoothly with the underlying metrics (for example, "Roughly 40% of customer negative sentiment is driven by return policy") rather than dumping raw tables.
* Invisible data mechanics: Ground your insights silently in the data. Never print a "Data Funnel," never show your filtering scratchpad, and never narrate backend tool execution steps.
* No unsolicited sections: Do NOT include "Suggested follow-ups," "Recommended actions," "Funnel breakdowns," or "Key conversations" sections unless the user explicitly asks for them.
* Clean and adaptive structure: Use bolding, concise paragraphs, and scannable bullet points organically based on what best answers the specific prompt.

# ANALYSIS PHILOSOPHY AND ZERO CLARIFICATION
You have total freedom to determine the best analytical path to answer the question. Decide for yourself which columns to inspect, whether to aggregate or enrich, and how to structure your findings.
NEVER ask the user for permission, parameter clarification, column names, filter logic, or output types. Make all analytical decisions behind the scenes, execute the data tools, and provide the synthesized answer.
```

## Things to know

**Streamable HTTP only.** Copilot Studio dropped SSE support for MCP, and Rippit speaks Streamable HTTP, so this is already correct, but it's why older MCP setup guides don't apply.

**Power Platform data policies apply.** Copilot Studio reaches MCP servers through Power Platform connectors, so any DLP policy that governs connectors also governs Rippit. If a locked-down environment blocks the connection, that policy is where to look.

**Microsoft 365 Copilot declarative agents are a separate path.** MCP plugins there are configured with the Microsoft 365 Agents Toolkit rather than the Copilot Studio wizard. The same endpoint and dynamic registration apply.

## Next steps

* [What Rippit can do for AI assistants](/mcp-server/overview#what-rippit-can-do-for-ai-assistants)
* [Troubleshooting](/troubleshooting)