MCP

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external tools and services. Sazabi uses MCP to extend Chat with capabilities from services like Slack, Linear, Datadog, and more.

How Sazabi uses MCP

Sazabi acts as an MCP client, connecting to MCP servers that expose tools. When you connect an MCP provider, Chat discovers the available tools and can use them during conversations.

┌─────────────┐      ┌────────────────┐      ┌──────────────┐
│   Sazabi    │ ──── │  MCP Protocol  │ ──── │  MCP Server  │
│   (Client)  │      │                │      │  (Provider)  │
└─────────────┘      └────────────────┘      └──────────────┘

For example, when you connect the Linear MCP connector:

1. Sazabi connects to Linear's MCP server
2. Linear exposes tools like create_issue, list_projects, update_status
3. Chat can use these tools when you ask it to interact with Linear

Available MCP connectors

Sazabi provides pre-configured connectors for popular services:

• Slack - Send messages, read channels, manage conversations
• Linear - Create issues, manage projects and cycles
• Datadog - Query metrics, create monitors
• incident.io - Manage incidents and runbooks
• Supabase - Query databases, manage auth
• Sentry - Access error data, manage issues

You can also connect custom MCP servers to integrate with internal tools or services that provide MCP endpoints.

Setting up connectors

To connect an MCP service:

1. Go to Settings -> MCP Connectors
2. Select a pre-configured connector or add a custom server
3. Complete the OAuth flow or provide authentication headers
4. Sazabi discovers and enables the available tools

For detailed setup instructions, see the MCP Connectors documentation.

Using MCP tools in Chat

Once connected, MCP tools work like any other Chat tool. Ask the assistant to perform actions in natural language:

Create a Linear issue for this bug with the stack trace
Post a summary to #incidents on Slack
Check for related incidents in incident.io

The assistant selects appropriate tools based on your request. You see each tool call as it happens, including the parameters used and results returned.

For details on how integration tools appear in conversations, see Integration Tools.

Next steps