OpenIndex Retrieval Plugin API MCP Server

1. Workflow Overview

This workflow, titled OpenIndex Retrieval Plugin API MCP Server, functions as a server endpoint for AI agents to perform natural language-based document retrieval and filtering via the OpenIndex Retrieval Plugin API. It exposes a single operation allowing AI agents to submit queries and receive responses directly, preserving the original API structure.

The workflow logic is divided into the following blocks:

1.1 Setup and Documentation: Contains detailed setup instructions and overview documentation for users to understand and deploy the workflow.
1.2 MCP Server Trigger: Listens for incoming AI agent requests on a webhook URL, acting as the server entry point.
1.3 API Request Handling: Processes incoming queries by forwarding them to the OpenIndex Retrieval Plugin API endpoint and returns the API responses to the AI agent.

2. Block-by-Block Analysis

1.1 Setup and Documentation

Overview:
This block provides users with comprehensive setup instructions, usage notes, customization tips, and support information. It ensures users can correctly import, configure, and operate the workflow.
Nodes Involved:
- Setup Instructions (Sticky Note)
- Workflow Overview (Sticky Note)
Node Details:
- Setup Instructions
  - Type & Role: Sticky Note node used for documentation.
  - Configuration: Contains a multi-section markdown text with setup steps, usage notes, customization advice, and support links.
  - Key Content: Instructions to import the workflow, activate it, retrieve the MCP URL, connect AI agents, and general customization guidelines. Includes a Discord support link and n8n docs for MCP nodes.
  - Inputs/Outputs: None (informational only).
  - Edge Cases: None, as it is purely descriptive.
- Workflow Overview
  - Type & Role: Sticky Note node summarizing workflow purpose and operation.
  - Configuration: Markdown content describing the workflow’s function as a retrieval API, operation details, the MCP trigger role, HTTP request handling, and parameter automation via AI expressions.
  - Inputs/Outputs: None (informational only).
  - Edge Cases: None.

1.2 MCP Server Trigger

Overview:
Acts as the webhook endpoint that listens for incoming requests from AI agents. This node initiates the workflow upon receiving a request.
Nodes Involved:
- OpenIndex Retrieval Plugin MCP Server (MCP Trigger)
Node Details:
- OpenIndex Retrieval Plugin MCP Server
  - Type & Role: MCP Trigger node, specialized for handling AI agent requests via a webhook.
  - Configuration:
    - Webhook Path: openindex-retrieval-plugin-mcp (defines the endpoint URL).
    - No authentication required (open endpoint).
  - Key Expressions: Parameters are dynamically populated from AI request context using $fromAI() expressions internally (not explicitly shown).
  - Input: Incoming webhook request from AI agent.
  - Output: Forwards data to the next node for processing.
  - Version-specific: Requires n8n version supporting MCP Trigger nodes.
  - Edge Cases: Possible failures include webhook misconfiguration, network errors, or malformed AI requests.

1.3 API Request Handling

Overview:
Handles the actual API interaction by submitting queries received from the MCP trigger to the OpenIndex Retrieval Plugin API and returning the raw API response to the AI agent.
Nodes Involved:
- Sticky Note ("Query")
- Submit Query (HTTP Request Tool)
Node Details:
- Sticky Note ("Query")
  - Type & Role: Sticky Note node labeling this block for clarity.
  - Configuration: Simple markdown header "## Query".
  - Inputs/Outputs: None.
- Submit Query
  - Type & Role: HTTP Request Tool node, performs the POST request to the API endpoint.
  - Configuration:
    - URL: https://retriever.openindex.ai/sub/query
    - Method: POST
    - Options: Default (no special headers or authentication).
    - Tool Description: "Query" (indicating the node’s functional role).
    - Parameters: Automatically populated from AI agent request via MCP context (through $fromAI() expressions handled internally by MCP trigger).
  - Input: Receives query data from MCP trigger node.
  - Output: Returns the API response directly back to the MCP node, which sends it to the AI agent.
  - Edge Cases:
    - Network timeouts or unreachable API endpoint.
    - Invalid or malformed query data causing API errors.
    - API rate limiting or service unavailability.
    - No authentication is required by the API, but if API changes, this might need update.

3. Summary Table

Node Name	Node Type	Functional Role	Input Node(s)	Output Node(s)	Sticky Note
Setup Instructions	Sticky Note	Provides setup and usage instructions	None	None	### ⚙️ Setup Instructions 1. Import Workflow; 2. No auth needed; 3. Activate workflow; 4. Get MCP URL; 5. Connect AI agent. Usage notes and customization tips included. Discord support link: https://discord.me/cfomodz and n8n docs link provided.
Workflow Overview	Sticky Note	Summarizes workflow purpose and usage	None	None	## 🛠️ OpenIndex Retrieval Plugin MCP Server: retrieval API, MCP trigger, HTTP request nodes, AI expressions, one operation supported.
OpenIndex Retrieval Plugin MCP Server	MCP Trigger	Entry point webhook for AI agent requests	None	Submit Query
Sticky Note ("Query")	Sticky Note	Labels the query processing block	None	None	## Query
Submit Query	HTTP Request Tool	Sends query to OpenIndex API and returns response	OpenIndex Retrieval Plugin MCP Server	None

4. Reproducing the Workflow from Scratch

Create Sticky Note: Setup Instructions
- Add a Sticky Note node.
- Paste detailed setup instructions: import steps, no auth needed, activation, MCP URL retrieval, AI agent connection, usage notes, customization tips, Discord link (https://discord.me/cfomodz), and n8n docs link (https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.toolmcp/).
- Set color to teal (color 4), height to approx. 1060.
Create Sticky Note: Workflow Overview
- Add another Sticky Note node.
- Enter markdown summarizing the workflow’s purpose as an OpenIndex Retrieval Plugin MCP server, describing the MCP trigger, HTTP requests, AI expressions, and supported operation.
- Set width to 420, height to 740.
Add MCP Trigger Node: OpenIndex Retrieval Plugin MCP Server
- Select the MCP Trigger node type (requires n8n version with Langchain MCP nodes).
- Set the webhook path to openindex-retrieval-plugin-mcp.
- No authentication required.
- This node will listen for incoming AI agent requests.
Add Sticky Note: Query Label
- Add a Sticky Note node with content "## Query".
- Set color to green (color 2), width 300, height 200.
Add HTTP Request Tool Node: Submit Query
- Add HTTP Request Tool node.
- Set Method to POST.
- Set URL to https://retriever.openindex.ai/sub/query.
- Leave options as default (no auth or headers).
- Enter tool description as "Query".
- Parameters will be auto-populated via MCP context from AI agent requests ($fromAI() expressions are managed internally by the MCP node).
Connect Nodes
- Connect MCP Trigger node output to Submit Query node input (using the ai_tool output).
- Connect Sticky Note "Query" nearby for visual grouping only.
Activate Workflow
- Enable workflow in n8n to start listening on the defined webhook URL.
Usage
- Provide the webhook URL (base n8n URL + /webhook/openindex-retrieval-plugin-mcp) to your AI agent.
- The AI agent sends queries, which are forwarded to the OpenIndex Retrieval Plugin API, with responses returned transparently.

5. General Notes & Resources

Note Content	Context or Link
No authentication is required for this workflow’s MCP endpoint or the OpenIndex API.	Workflow Setup Instructions
Parameters in HTTP requests are auto-populated by AI agent context using `$fromAI()` expressions internally.	Workflow Overview and Setup Instructions
Discord support channel available for integration help: https://discord.me/cfomodz	Setup Instructions
Official n8n documentation for MCP and Langchain nodes: https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.toolmcp/	Setup Instructions
The workflow exposes a single operation endpoint: Submit Query.	Workflow Overview

Disclaimer: The provided content originates solely from an automated n8n workflow. All data handled is legal and public, adhering to content policies.

11 KiB Raw Blame History Unescape Escape