Custom MCP Clients

Goal: Build your own MCP client to query OpenIndustrial data.

---

MCP Protocol Basics

MCP uses JSON-RPC 2.0 over stdio or HTTP:

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "query_temperature_avg",
    "arguments": {
      "timeRange": "1h"
    }
  },
  "id": 1
}

Response:

{
  "jsonrpc": "2.0",
  "result": {
    "content": [
      {
        "type": "text",
        "text": "{\"avg_temp\": 72.5, \"min\": 68.2, \"max\": 76.1}"
      }
    ]
  },
  "id": 1
}

---

Option 1: Use the SDK

Install the OpenIndustrial MCP SDK:

npm install @openindustrial/mcp-client

TypeScript Example

import { OpenIndustrialMCP } from '@openindustrial/mcp-client';

const client = new OpenIndustrialMCP({
  workspaceId: 'your-workspace-id',
  token: process.env.OI_API_TOKEN
});

// List available tools
const tools = await client.listTools();
console.log('Available queries:', tools.map(t => t.name));

// Call a tool
const result = await client.callTool('query_temperature_avg', {
  timeRange: '1h'
});
console.log('Result:', result);

Python Example

from openindustrial import MCPClient

client = MCPClient(
    workspace_id='your-workspace-id',
    token=os.environ['OI_API_TOKEN']
)

# List available tools
tools = client.list_tools()
print('Available queries:', [t.name for t in tools])

# Call a tool
result = client.call_tool('query_temperature_avg', {'timeRange': '1h'})
print('Result:', result)

---

Option 2: Direct HTTP

If you can't use the SDK, call the MCP HTTP endpoint directly:

List Tools

curl -X POST https://mcp.openindustrial.co/{workspace-id}/rpc \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc": "2.0", "method": "tools/list", "id": 1}'

Call Tool

curl -X POST https://mcp.openindustrial.co/{workspace-id}/rpc \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "query_temperature_avg",
      "arguments": {"timeRange": "1h"}
    },
    "id": 2
  }'

---

Option 3: Stdio Transport

For subprocess-based integrations:

npx @openindustrial/mcp-server \
  --workspace YOUR_WORKSPACE_ID \
  --token YOUR_TOKEN

Send JSON-RPC over stdin, receive over stdout.

---

Tool Discovery

Every warm query exposes its schema:

{
  "name": "query_temperature_avg",
  "description": "Get average temperature for a time range",
  "inputSchema": {
    "type": "object",
    "properties": {
      "timeRange": {
        "type": "string",
        "description": "Time range (e.g., '1h', '24h', '7d')"
      },
      "deviceId": {
        "type": "string",
        "description": "Optional device filter"
      }
    },
    "required": ["timeRange"]
  }
}

---

Building an AI Agent

Example: Autonomous monitoring agent

import { OpenIndustrialMCP } from '@openindustrial/mcp-client';
import { ChatOpenAI } from 'langchain/chat_models/openai';
import { AgentExecutor } from 'langchain/agents';

// Connect to OpenIndustrial
const mcp = new OpenIndustrialMCP({
  workspaceId: 'your-workspace-id',
  token: process.env.OI_API_TOKEN
});

// Get tools as LangChain format
const tools = await mcp.asLangChainTools();

// Create agent
const agent = AgentExecutor.fromAgentAndTools({
  agent: new ChatOpenAI({ model: 'gpt-4' }),
  tools: tools
});

// Run autonomous monitoring
const result = await agent.invoke({
  input: 'Check all temperature sensors and alert if any are above threshold'
});

---

Error Handling

MCP errors follow JSON-RPC conventions:

{
  "jsonrpc": "2.0",
  "error": {
    "code": -32600,
    "message": "Invalid Request",
    "data": {"detail": "Missing required parameter: timeRange"}
  },
  "id": 1
}

Code	Meaning
-32600	Invalid request
-32601	Method not found
-32602	Invalid params
-32603	Internal error
-32001	Authentication failed
-32002	Authorization denied

---

Rate Limiting

MCP endpoints are rate-limited:

Tier	Requests/minute
Standard	60
Professional	300
Enterprise	Custom

Handle 429 responses with exponential backoff.

---

Next Steps

If you want to...	Go to...
Use pre-built integrations	App Makers →
Connect Claude Desktop	Claude Desktop →
Manage API keys	Secrets →