PLCs.ai MCP Server

Connect Claude, Cursor, Codex, or any MCP-compatible agent to your PLC projects.

PLCs.ai is the AI platform purpose-built for PLC engineers: upload an Allen-Bradley (Studio 5000), Siemens (TIA Portal) or CODESYS V3 project and understand, troubleshoot, document, and safely generate code for it in plain English. This MCP server gives any AI agent governed access to that same capability — your projects, live tag values, and analyses — with per-tool permission scopes.

Remote server: https://mcp.plcs.ai/mcp (Streamable HTTP · OAuth or API key) Full documentation: https://www.plcs.ai/developer/mcp

Quick start

Claude Code

claude mcp add --transport http plcs-ai https://mcp.plcs.ai/mcp

Then run /mcp and choose Authenticate to sign in with your PLCs.ai account. (Or pass an API key: --header "Authorization: Bearer plck_live_…".)

Claude Desktop

Settings → Connectors → Add custom connector → URL https://mcp.plcs.ai/mcp → Connect and sign in. For API-key or on-prem setups, use the mcp-remote bridge:

{
  "mcpServers": {
    "plcs-ai": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.plcs.ai/mcp",
               "--header", "Authorization: Bearer plck_live_…"]
    }
  }
}

Cursor / Codex / other MCP clients

See per-client guides: https://www.plcs.ai/developer/mcp — covers Cursor, Codex, and any client that speaks remote MCP or can run mcp-remote.

Related MCP server: OT-AIops

What the agent can do

17 tools, permission-scoped per API key or OAuth grant:

The agent reasons over projects your organization already owns. Onboarding a machine is a deliberate act in the app or through a version-source connector, so there is no create-a-project tool — an agent can read, ask, propose and save versions, but never add a project.

Example prompts

• "List my PLC projects and check which ones have live data connected."

• "Why won't Conveyor 3 restart after a stop? Trace the interlocks."

• "Read the latest analysis on the Palletizer project and summarize what's safe to remove."

• "What did tag Motor1_Enable do in the last hour?"

• "Draft a fix for the missing handshake between Conv001 and Rob010 — proposal only."

Safety model

• Every tool call is authenticated and scoped to your organization; the credential is the tenant.

• Proposing code is two steps — a plan you approve, then the change — and it returns a proposal either way; persisting a new version is a separate, separately-scoped action.

• Customer data is never used to train AI models. Data handling →

Requirements

• A PLCs.ai account (start a free trial) or API key

• For the mcp-remote bridge: Node 18+

© PLCs.ai · Terms · Privacy · Questions: sales@plcs.ai