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

# 🤖 MCP Integration

> Connect Orsay to Claude and any MCP-compatible client to manage sequences and conversations in plain English

The **Model Context Protocol (MCP)** is an open standard that lets AI clients (Claude, Cursor, etc.) call external tools directly from a prompt. Orsay exposes its data — conversations, sequences, analytics, leads — through a secure, authenticated MCP server.

The result: query your pipeline in plain English, trigger Orsay actions, and chain multiple tools together (Notion, Linear, Gmail…) — no code required.

## Connect

**1. Open Connect MCP in Orsay**

Go to **[Integrations](https://app.orsay.ai/integrations)** and click **[Connect MCP](https://app.orsay.ai/integrations/claude-mcp/connect-claude-mcp)**.

You'll find the MCP server URL to copy:

```
https://client.api.prod.orsay.ai/mcp
```

**2. Paste the URL into your MCP client**

Open your MCP client (Claude.ai, Claude Desktop, Cursor…) and add a new server with that URL. See the [Compatible clients](#clients) section below for step-by-step instructions.

**3. Authorize Orsay**

Your client will redirect you to Orsay to sign in and grant access. Authorization is **scoped to your organization** — no other account can read your data.

## Compatible clients

> [tab] Claude.ai (web)
>
> 1. Go to **[claude.ai/settings/connectors](https://claude.ai/settings/connectors)**
> 2. Click **Add custom connector**
> 3. Paste the URL: `https://client.api.prod.orsay.ai/mcp`
> 4. Give it a name (e.g. *Orsay*) and click **Add**
> 5. Authorize the connection in the Orsay window that opens
>
> Once connected, the Orsay icon appears in Claude's toolbar and the tools are available in all your conversations.

> [tab] Claude Desktop
>
> Open your Claude Desktop config file:
>
> * **macOS**: `~/Library/Application Support/Claude/claude_desktop_config.json`
> * **Windows**: `%APPDATA%\Claude\claude_desktop_config.json`
>
> Add the Orsay server under `mcpServers`:
>
> ```json theme={null}
> {
>   "mcpServers": {
>     "orsay": {
>       "url": "https://client.api.prod.orsay.ai/mcp"
>     }
>   }
> }
> ```
>
> Restart Claude Desktop. Orsay will appear in the available tools list.

> [tab] Cursor
>
> 1. Open **Settings** → **MCP**
> 2. Click **Add new global MCP server**
> 3. Paste the URL: `https://client.api.prod.orsay.ai/mcp`
> 4. Save and restart Cursor if prompted

> [tab] Other MCP clients
>
> Any client supporting the MCP protocol (Streamable HTTP) can connect using:
>
> ```
> https://client.api.prod.orsay.ai/mcp
> ```
>
> Refer to your client's documentation for how to add a custom MCP server.

## Available tools

13 tools are exposed. You don't need to call them directly — Claude selects and chains them automatically based on your prompt.

### 💬 Conversations & leads

| Tool | What it does |
| - | - |
| `list_conversations` | List conversations with optional filters: classification (positive, negative, engaged…), channel (WhatsApp / Instagram), result count |
| `get_conversation` | Retrieve the full thread of a conversation with all its messages |
| `search_conversations` | Search across conversations by text content |
| `get_lead_info` | Lead details: name, channel, creation date, profile picture |

### 📊 Analytics

| Tool | What it does |
| - | - |
| `get_analytics` | Metrics over a date range: contacts reached, classification breakdown (positive, engaged, negative, undetermined), funnel (conversations started, qualified, links sent). Filterable by sequence or channel |

### ⚡ Sequences

| Tool | What it does |
| - | - |
| `list_sequences` | List all sequences in the organization (name, trigger, channel, active status) |
| `get_sequence_detail` | Full details of a sequence: flow steps, associated agent, working hours, advanced settings |
| `create_sequence` | Create a new sequence: trigger, message flow, agent, response delay, working hours… |
| `update_sequence` | Update an existing sequence — only provided fields are changed |
| `delete_sequence` | Permanently delete a sequence |

### 🤖 Agents

| Tool | What it does |
| - | - |
| `list_agents` | List AI agents available in the organization |
| `list_filtering_agents` | List filtering agents (lead qualification before engagement) |

### 🏢 Organization

| Tool | What it does |
| - | - |
| `list_organizations` | List all organizations your account belongs to — useful if you manage multiple workspaces |

## Prompt examples

Ready-to-use prompts for Claude.

**Analyze your pipeline**

> *"How many leads replied positively in my Instagram sequences this month? Compare with last month."*

**Ad-hoc segmentation**

> *"Find leads who mentioned 'pricing' in the last 30 days but haven't received my pricing sequence yet."*

**Natural-language reporting**

> *"Compare the positive classification rates of my 3 active sequences over the past 14 days and tell me which one performs best."*

**Summarize a conversation**

> *"Summarize the conversation with lead \[leadId] and write a personalized follow-up based on their last replies."*

**Create a sequence**

> *"Create an Instagram inbound sequence for the account @myaccount using the 'Sales Agent' agent, a 2-minute response delay, and a welcome message 'Hi there, how can I help you?'. Leave it inactive for now."*

**Update a sequence**

> *"Activate the 'Spring 2026 Campaign' sequence and change its response delay to 5 minutes."*

**Cross-tool workflow** (combining other MCP servers)

> *"Read this week's positive Orsay conversations, create a Linear ticket for each one with the lead context, then generate a summary in Notion."*

## Technical reference

> [accordion] Available triggers for create\_sequence
>
> | Value | Description |
> | - | - |
> | `CONTACT_CREATED` | New contact created in the organization |
> | `CONTACT_SUBSCRIBED_TO_SEQUENCE` | Contact manually added to a sequence |
> | `WHATSAPP_MESSAGE` | Incoming WhatsApp message (inbound) |
> | `INSTAGRAM_COMMENT` | Comment on an Instagram post |
> | `INSTAGRAM_MESSAGE` | Incoming Instagram direct message |
> | `INSTAGRAM_STORY_REPLY` | Reply to an Instagram story |

> [accordion] Delay units
>
> `seconds` · `minutes` · `hours` · `days`

> [accordion] Classifications
>
> | Value | Meaning |
> | - | - |
> | `positive` | Qualified lead, interested |
> | `engaged` | Lead replied but not yet fully qualified |
> | `negative` | Unqualified or uninterested lead |
> | `undetermined` | No reply yet |

## Important notes

> [!WARNING]
>
> The `create_sequence`, `update_sequence`, and `delete_sequence` tools **modify your Orsay account in real time**. Always review Claude's proposed actions before confirming.

> [!TIP]
>
> If you manage multiple organizations, Claude will automatically call `list_organizations` to ask which one to use. Mention the organization name in your prompt to skip that step.

* Access is authenticated via OAuth 2.0 — only your organization's data is accessible.
* The MCP server is **stateless**: each request is independent, with no persistent server-side session.
* Compatible with any client supporting the MCP protocol over Streamable HTTP.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.