Connecting Your Tools (MCP)

Sire is built on the Model Context Protocol (MCP)---the universal standard for tool communication. With MCP, you can connect almost any software or service to Sire without writing custom integrations.

What is an MCP Server?

An MCP server is a lightweight application that "exposes" tools to Sire. For example, a Playwright MCP server exposes browser automation tools, while a Slack MCP server exposes messaging tools.

Registering Your First MCP Server

1. Navigate to MCP Registry: In Mission Control, go to Settings > MCP Servers.
2. Add New Server: Click Register Server.
3. Provide Connection Details:
   • Name: Give it a short name (e.g., playwright).
   • URL: Enter the HTTP endpoint where the MCP server is hosted (e.g., http://localhost:3000/rpc).
4. Confirm Connection: Sire will check the endpoint and verify its compatibility.

Using MCP Tools in Your Workflow

Once registered, Sire's AI will automatically "learn" about the tools provided by that server. When you prompt a mission, the AI will use the new tools as needed.

Standard URI Syntax

You can also manually reference any MCP tool using our standard URI scheme:

• mcp:server-name#tool-name
  • Example: mcp:playwright#browser.navigate
  • Example: mcp:github#repo.create_issue

Auto-Resolution

Sire automatically handles the communication with these servers, including:

• Parameter Passing: Mapping your workflow data to the tool's expected inputs.
• Result Handling: Parsing the JSON response and making it available to downstream steps.
• Error Management: Retrying failed calls and reporting timeouts.

Built-In Tools

Sire also comes with a suite of "native" tools that don't require an external server. These are highly optimized and available to every workflow:

• sire:local/http.request: Call any REST API directly.
• sire:local/file.read & write: Interact with a sandboxed file system.
• sire:local/logic.if & switch: Create conditional branches in your DAG.
• sire:local/data.transform: Use simple expressions to filter, map, or modify JSON data.

Why MCP?

No Vendor Lock-In

Because MCP is an open standard, you aren't tied to Sire's internal plugins. You can move your MCP servers between different platforms or write your own in any language.

Specialized AI Capabilities

You can build or buy specialized MCP servers for niche tasks---like bioinformatics, legal analysis, or industrial IoT---and they will plug into Sire instantly.

Security First

MCP servers can be hosted within your private network, ensuring that sensitive data and powerful tools never leave your control while being orchestrated by Sire.

Pre-Built MCP Servers

Sire provides a marketplace of ready-to-use MCP servers for common integrations.

Enabling a Pre-Built Server

1. Go to Settings > MCP Servers.
2. Browse the Marketplace tab.
3. Click Enable on the server you want.
4. Add any required credentials (API keys, OAuth tokens) when prompted.

Building a Custom MCP Server

When you need to connect a tool or service that is not in the marketplace, you can build your own MCP server. This tutorial walks through the process.

Option A: Generate with Mint (Recommended)

If your service has an OpenAPI specification, use Sire's open-source Mint CLI to generate a complete MCP server automatically.

# Install Mint
go install github.com/sirerun/mint@latest

# Generate an MCP server from your OpenAPI spec
mint mcp generate --spec your-service-openapi.yaml --output ./my-server

# Validate the generated server
mint validate ./my-server

# Build
cd my-server
go build -o server .

Mint generates a complete Go project with:

• JSON-RPC 2.0 request handling
• Tool definitions derived from your OpenAPI endpoints
• Input validation
• A Dockerfile for container deployment

Option B: Build Manually

If your service does not have an OpenAPI spec, implement the MCP protocol directly. Your server must handle two JSON-RPC methods.

1. Implement tools/list

This method returns the tools your server provides.

// Request
{"jsonrpc": "2.0", "method": "tools/list", "id": 1}

// Response
{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "tools": [
      {
        "name": "get_weather",
        "description": "Get the current weather for a city",
        "inputSchema": {
          "type": "object",
          "properties": {
            "city": {"type": "string", "description": "City name"}
          },
          "required": ["city"]
        }
      }
    ]
  }
}

2. Implement tools/call

This method executes a tool with the given parameters.

// Request
{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "id": 2,
  "params": {
    "name": "get_weather",
    "arguments": {"city": "San Francisco"}
  }
}

// Response
{
  "jsonrpc": "2.0",
  "id": 2,
  "result": {
    "content": [
      {"type": "text", "text": "San Francisco: 62F, partly cloudy"}
    ]
  }
}

Deploy Your Server

Host your MCP server anywhere that Sire can reach it:

• Cloud Run / Cloud Functions -- serverless, scales to zero
• Kubernetes -- for servers that need persistent connections
• Your private network -- Sire can connect via tunnel for on-premise services

Register and Test

1. Go to Settings > MCP Servers and click Register Server.
2. Enter the server name and URL.
3. Click Test Connection -- Sire will call tools/list to verify compatibility.
4. Once registered, your tools appear in the workflow editor and are available to the AI when generating workflows.

Troubleshooting MCP Connections