Skip to main content

Legacy AI Chain Deployment (migrating to Tools in Conversation Mode)

Legacy reference

This page documents the integration UX for the now-deprecated AI Chains primitive. AI Chains have been unified into Tools with Conversation mode (per EN-3122). Existing AI Chains continue to work, and their API/integration bindings continue to function unchanged thanks to ID preservation — see Migration from AI Chains.

For new development, use the standard Deployment tab on a Tool (with Conversation mode toggled on if needed). This page is preserved for historical reference only.

Use the Integrations tab to connect a AI chain to external apps and clients. AI chain integrations provide three sub-tabs:

  • MCP Configuration – expose one or more AI chains behind a SimplAI-hosted MCP server.
  • API Configuration – call your AI chain over HTTPS.
  • Webhooks – receive asynchronous results via POST callbacks.

Before you begin​

  • Your AI chain must have at least one Published version.
  • The first time you open Integrations after publishing, click Enable Integration. If you later disable it, the button will appear again while a Published version exists.
  • Depending on your plan, enabling integrations for AI chains may count against your organization’s integratable resources quota.

MCP Configuration​

Create a managed MCP server that can host multiple AI chains and expose them to MCP-compatible clients (e.g., Cursor, Claude Desktop, custom SSE clients).

Create a server

  1. Go to Integrations → MCP Configuration and click Create New.
  2. Enter a Name and Description.
  3. Select one or more AI chains to host.
  4. Click Generate MCP server configuration.

Use the configuration

  • The dialog shows a JSON snippet containing the SSE URL hosted by SimplAI and an auth token embedded in the URL.
  • Copy this into your MCP client configuration. The same URL works for Cursor and any other platform that accepts an SSE URL with token.
  • Each server card lists the selected AI chains and creation time.
    • View (eye) to re-open the configuration JSON.
    • Edit (pencil) to change included AI chains.
    • Delete (trash) to remove the server.

Notes

  • A AI chain must be Published to be callable through the MCP server.
  • Treat the URL as a secret; rotate by regenerating the server or your API key.

API Configuration​

Call a AI chain directly via REST.

Credentials shown

  • API Key, AI chain ID, Tenant ID, Client ID.

Run your AI chain

  • Endpoint: POST /interact/api/v1/run/AI chain
  • Headers: PIM-SID, X-TENANT-ID, X-CLIENT-ID (plus standard auth)

Body:
{
"AI chain_id": "<AI chain-id>",
"language_code": "EN",
"inputs": {
"yourInputKey1": "value1",
"yourInputKey2": 42
}
}

  • Response returns conversation_id and message_id for polling.

Fetch results

  • Endpoint: GET /interact/api/v1/run/AI chain/result
  • Include conversation_id and message_id as query params.
  • When message_status indicates completion, parse the output payload.

An example curl is provided in the UI under Example Request.

Webhooks​

Receive AI chain results without polling.

Configure

  1. Open Integrations → Webhooks.
  2. Choose a Webhook Event and enter an HTTPS URL.
  3. Click Add Webhook.

Usage

  • Use polling when your client can periodically check status.
  • Use webhooks to have SimplAI push results automatically to your endpoint.

Troubleshooting​

  • Integrations options are disabled: Publish the AI chain, then click Enable Integration.
  • AI chain not visible in MCP clients: Ensure the AI chain is Published and included in the selected MCP server; re-copy the config URL with token.
  • 401/403 from API: Verify API key and required headers; confirm Tenant/Client IDs.
  • Webhook not firing: Confirm the event is added, the URL is HTTPS and reachable (returns 2xx), and your firewall allows inbound requests.

Security​

  • Keep API keys and MCP URLs with tokens confidential.
  • Regenerate keys/servers if exposed.
  • Limit access to only the AI chains you intend to host on a given MCP server.