# OpenAI MCP > OpenAI MCP manages your entire AI resource stack conversationally. List and track all models, monitor fine-tuning jobs, manage Assistants, and run cost-effective batch processing—all without leaving your agent's chat window. ## Overview - **Category:** ai-frontier - **Price:** Free - **Tags:** llm-management, fine-tuning, model-discovery, ai-assistants, file-management, api-orchestration ## Description Connecting your OpenAI account to your agent means you get full oversight of your model infrastructure through natural conversation. Instead of jumping between dashboards just to check job statuses or audit resources, you talk to your AI client. You can discover every available model, from GPT-4o to DALL-E 3, and pull up its ownership details. Need to stop a bad training run? The MCP lets you monitor fine-tuning jobs and cancel them instantly. It also handles file management for all your uploaded data, making it simple to delete old or unused assets. For bulk work, you can create batch processing jobs to handle hundreds of API calls cost-effectively. You'll find managing model lifecycles much easier when connected through Vinkius, letting your agent act as a dedicated ML ops assistant. ## Tools ### cancel_batch Stops an active batch processing job using its unique ID. ### cancel_fine_tune Halts a running fine-tuning job when you need to restart or change the source material. ### create_batch Initiates a new, large-scale batch processing job using specific input files and endpoints. ### delete_file Permanently removes an uploaded file asset from your account. Be careful with this one. ### get_assistant Retrieves detailed information about a specific Assistant by its ID. ### get_batch Pulls the current status and details for a specified batch job. ### get_fine_tune Checks the detailed status, base model, and progress of a fine-tuning job ID. ### get_model Verifies if a specific OpenAI model exists and retrieves its metadata (e.g., owner or creation date). ### list_assistants Generates an audit list of all configured Assistants, detailing their models and tools. ### list_batches Provides a status overview for every batch processing job in your account. ### list_files Shows all uploaded files, noting if they are intended for fine-tuning, Assistants, or batches. ### list_fine_tunes Lists the status of all your model training jobs to monitor the overall pipeline health. ### list_models Displays every available OpenAI model ID and its capability flags for discovery. ## Prompt Examples **Prompt:** ``` Show me all available GPT models. ``` **Response:** ``` I found 12 GPT models in your account: gpt-4o (owned by openai, created 2024-05), gpt-4o-mini (lightweight, cost-effective), gpt-4-turbo, gpt-4, gpt-3.5-turbo, plus several fine-tuned variants. Would you like details on any specific model? ``` **Prompt:** ``` Check the status of my latest fine-tuning job. ``` **Response:** ``` Your latest fine-tune job ftjob-abc123 is currently 'running' at 67% completion. It's training on 1,240 examples using gpt-3.5-turbo as the base model. Estimated finish is in approximately 2 hours. The training file is file-xyz789. ``` **Prompt:** ``` List all my uploaded files and their purposes. ``` **Response:** ``` You have 8 files uploaded: 3 for fine-tuning (training JSONL files), 2 for Assistants (knowledge base PDFs), 2 for batch processing (request JSONL files) and 1 vector store file. Total storage used: 45MB. ``` ## Capabilities ### Audit Model Inventory Discover all models available to your account and check their metadata, like ownership or creation date. ### Manage Fine-Tuning Pipelines Monitor the status of training jobs, track progress, and cancel long-running fine-tuning processes. ### Control Assistant Configurations List and inspect all configured Assistants, checking their instructions, models, and tools before deployment. ### Process Bulk API Requests Set up and track batch processing jobs to run large volumes of requests cost-effectively. ### Handle File Assets List, manage, and delete uploaded files used for fine-tuning or Assistants. ## Use Cases ### The Model Discovery Check A product manager needs to know if the newest embedding model is available. Instead of reading through documentation or hitting an API endpoint manually, they ask their agent: 'What models do I have?' The agent runs list_models and reports back a clean list with capabilities. ### Stopping Wasteful Training An ML engineer realizes they uploaded the wrong training data. Rather than waiting hours for a job to fail, they ask their agent to check the jobs using list_fine_tunes and then immediately execute cancel_fine_tune on the incorrect run ID. ### Auditing Assistant Dependencies A developer needs to verify which models are attached to a key customer-facing bot. They ask their agent to list_assistants, getting an instant report detailing every linked model and tool for audit purposes. ## Benefits - Stop jumping between dashboards. You manage model versions, fine-tuning jobs, and Assistant configurations entirely through chat with your agent. - Save time on bulk processing. Instead of manually submitting requests, you use create_batch to run large API workloads cost-effectively. - Maintain a clean workspace. Use list_files and delete_file to audit and remove old or unnecessary training data and assets. - Quickly diagnose problems. If an Assistant isn't working right, you can get_assistant to inspect its model, tools, and instructions immediately. - Track every job status in one place. From monitoring a running fine-tuning process with get_fine_tune to checking list_models for availability, it’s all conversational. ## How It Works The bottom line is you manage complex AI operations using simple conversational prompts instead of navigating multiple web dashboards. 1. Subscribe to this MCP on Vinkius and provide your OpenAI API Key. 2. Your AI client connects the key and pulls a real-time view of all your existing resources (models, files, etc.). 3. You ask your agent to perform an action—like listing assistants or canceling a batch job—and it executes the command directly. ## Frequently Asked Questions **How do I check if a specific model is available using OpenAI MCP?** You use get_model to verify existence and pull metadata. This tool confirms if the model ID you need—like 'gpt-4o'—is active in your account before you write any code that depends on it. **What is the difference between list_files and list_models?** list_files tracks data assets (PDFs, JSONL files) used for training or Assistants. list_models tracks the actual AI model types themselves (e.g., Whisper-1). You need both to run a full pipeline. **Can I use OpenAI MCP to stop an expensive fine-tune job?** Yes, you can. First, check status with list_fine_tunes, then execute cancel_fine_tune using the specific job ID. This stops unnecessary spending immediately. **How do I manage my Assistants through the OpenAI MCP?** You start by listing all assistants with list_assistants to get an overview. Then, you use get_assistant if you need deep details on one specific bot's configuration. **Is batch processing done via OpenAI MCP safe for large volumes?** Yes, create_batch is designed for this. It handles the workload of thousands of requests in a cost-effective way, and you track its status using list_batches.