Quickstart
Connect in a minute.
Point any MCP-capable client at the server URL and authenticate. Pick your client below for exact steps.
1. Choose how you'll authenticate
| Method | Best for |
|---|---|
| Cloud OAuth — sign in with your PLCs.ai account in the browser; the client stores a short-lived token. | Desktop & IDE clients on your own machine. |
API key — paste a plck_live_… bearer into the client config. | Headless / server setups and on-prem. |
Both are first-class — details in Authentication.
2. Add the server
The server speaks Streamable HTTP at:
text
https://mcp.plcs.ai/mcpMost clients accept either a bare URL (OAuth) or a URL plus an Authorization header (API key). This generic shape works in any client that reads an mcpServers map (Cursor, Claude Code, and more):
json
{
"mcpServers": {
"plcs-ai": {
"type": "http",
"url": "https://mcp.plcs.ai/mcp"
}
}
}json
{
"mcpServers": {
"plcs-ai": {
"type": "http",
"url": "https://mcp.plcs.ai/mcp",
"headers": { "Authorization": "Bearer plck_live_…" }
}
}
}3. Per-client setup
| Client | Guide |
|---|---|
| Claude Desktop | Add a custom connector → |
| Claude Code (CLI) | claude mcp add → |
| Codex | ~/.codex/config.toml → |
| Cursor | mcp.json → |
4. Verify
Ask your assistant to call plcs_health (or just "list my PLCs.ai projects"). A healthy response returns your organization id and actor type — you're connected.
On-prem? Replace the URL with your appliance's http://APPLIANCE_HOST:8080/mcp. Everything else is identical.