# Unstructured MCP for AI Agents AI Agent Connect > Unstructured MCP for AI Agents. This Connector lets you manage data ingestion pipelines, trigger document processing jobs, and monitor RAG data workflows directly from your AI client. It connects your agent to Unstructured.io so you can handle messy file transformations without leaving your editor. ## Overview - **Category:** ai-frontier - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_wLKl0zBun6GGEWOAXV2AtIEfq4BCwiQMTDSCBvt6/ai-agent-connect - **Tags:** rag, data-ingestion, document-processing, etl, unstructured-data, pipeline-automation ## Description You're trying to get messy documents into a clean vector database, but the manual work of cleaning and partitioning files is killing your momentum. It's a chore to switch between your code editor and a web dashboard just to check if a job finished or to see which sources are connected. This Connector gives your AI agent a direct line to Unstructured.io so you can manage the whole pipeline without leaving your workspace. Instead of digging through a UI to see if a job failed, you just ask your agent. It can pull data from S3 or SharePoint, send it to your SQL database, and tell you exactly where a job failed. Since Vinkius hosts this Connector, you can get these data engineering capabilities into Claude, Cursor, or Windsurf in minutes. It turns your agent into a hub for your RAG ingestion, letting you trigger new jobs or audit existing workflows while you stay focused on building the actual product. You can check your data destinations, list all your active sources, and see the status of every active job without ever opening a browser tab. It takes the friction out of document processing. This is about moving from manual data cleaning to automated pipelines where your agent handles the status checks and job triggers for you. ## Tools ### get_workflow_details Retrieves configuration details for a specific processing workflow. Use it to check exact source and destination rules. ### trigger_workflow_execution Manually triggers an immediate execution of a processing workflow. It returns a job ID so you can track the progress. ### list_data_destinations Lists all configured target locations for processed data. Use this to confirm your Vector DB or SQL links are active. ### list_processing_workflows Lists all end-to-end document processing pipelines. This helps you see every ingestion path you have set up. ### list_workflow_jobs Lists all active and historical workflow execution jobs. Use this to quickly see if your latest ingestion finished. ### list_data_sources Lists all configured remote data connectors. This lets you see all active S3, GCS, or SharePoint connections. ## Prompt Examples **Prompt:** ``` Show me all our active destination connectors. ``` **Response:** ``` You have 3 active destinations configured in Unstructured. | Destination | Type | Purpose | | :--- | :--- | :--- | | **Pinecone Index** | Vector DB | Production Knowledge Base | | **MongoDB Atlas** | Vector DB | Search Cluster | | **AWS S3** | Storage | Raw JSON Output | Would you like me to check which workflows are currently sending data to the Pinecone index? ``` **Prompt:** ``` List the historical processing jobs from today. ``` **Response:** ``` I found 2 workflow jobs executed today: * **Job ID `wf_92jdfk`**: Completed successfully at 08:30 AM (Ingested 450 PDFs from Sharepoint). * **Job ID `wf_44klqp`**: Failed at 11:15 AM (Error connecting to destination Pinecone timeout). Would you like me to share more log details about the failed job? ``` **Prompt:** ``` Trigger the engineering onboarding workflow. ``` **Response:** ``` I have successfully triggered the workflow `wf_eng_onboarding`. * **Job ID**: `job_12bxc6` * **Status**: Started * **Source**: Google Drive It is currently processing files. Do you want me to monitor it and let you know when it's populated into the Vector DB? ``` ## Capabilities ### Trigger document jobs Start a new ingestion job immediately and get a job ID for tracking. ### List data sources See all your configured remote connectors like S3 or SharePoint in one list. ### Monitor job status Check the progress of active and historical document processing tasks. ### View workflow details Pull the specific configuration and rules for any individual pipeline. ### Check data destinations Verify where your structured data is being sent, such as to a Vector DB or SQL. ### List processing workflows See every end-to-end pipeline you have configured for document ingestion. ## Use Cases ### Automated RAG updates A developer wants to sync SharePoint docs to a Vector DB. They ask the agent to trigger the workflow and then use list_workflow_jobs to confirm it finished. ### Onboarding document processing An HR tech lead needs to process 500 engineering PDFs. They use trigger_workflow_execution to start the batch and monitor it in real time. ### Source auditing A data engineer needs to see all active S3 connections. They ask the agent to list_data_sources to verify the correct buckets are linked. ### Pipeline debugging A job failed in the middle of the night. The MLOps engineer uses get_workflow_details to check the configuration and find the error. ## Benefits - Stop clicking through dashboards by using list_workflow_jobs to check status directly in your editor. - Speed up RAG setup by using list_data_sources to see exactly what feeds are ready for ingestion. - Debug failed pipelines faster with get_workflow_details to see specific configuration errors. - Automate document ingestion by using trigger_workflow_execution to start jobs via voice or text. - Keep your vector database updated without manual work by using list_data_destinations to verify targets. - Audit your entire data architecture by using list_processing_workflows to see every active pipeline. ## How It Works The bottom line is you get a command center for your RAG data pipelines inside your existing AI workflow. 1. Subscribe to the Connector and provide your Unstructured API Key and URL. 2. Connect your preferred AI client like Cursor or Claude to the Vinkius catalog. 3. Ask your agent to list sources, trigger jobs, or check pipeline statuses. ## Frequently Asked Questions **Can I use the Unstructured MCP to manage my RAG data pipelines?** Yes. It gives your AI agent the ability to trigger jobs, check statuses, and manage sources for your RAG system directly from your workspace. **Does the Unstructured MCP support S3 and SharePoint?** Yes, it allows your agent to interact with various remote data connectors including S3, GCS, and SharePoint to pull in documents for processing. **How do I check if my document processing job finished?** You can just ask your agent to list the workflow jobs. It will show you the status of both active and completed tasks with their specific IDs. **Can I trigger a new ingestion job using this Connector?** Absolutely. You can tell your agent to start a specific workflow immediately, and it will trigger the job and provide you with a tracking ID. **Is this Connector good for handling large PDF files?** Yes, it's designed specifically to manage the ingestion and partitioning of complex files like PDFs, Word docs, and HTML into clean data. **What happens if a data job fails using the Unstructured MCP?** Your agent can retrieve the specific configuration details for that workflow, helping you identify whether the issue is with the source, the destination, or the processing rules. **Can my AI agent trigger an immediate document processing job?** Yes! If you have a workflow configured to pull files from an S3 bucket and load them into a Pinecone index, you can ask your agent to `trigger workflow XYZ`. It will start the execution and return the new Job ID, which you can use to track the progress. **How can I verify if my RAG pipelines are failing or succeeding?** Ask your agent to list your workflow jobs. It will securely connect to Unstructured's engine and return historical and active executions, displaying statuses such as 'completed', 'failed', or 'in_progress'. This is extremely useful for MLOps engineers diagnosing ingestion alerts directly in their terminal. **Can I edit the destination database directly through the agent?** This server is focused on auditing and executing your existing pipelines. Currently, you can list all connections (sources and destinations) and obtain their details, but creating or destructively modifying vector database connectors must be done inside the Unstructured dashboard for security.