# OpenAI MCP for resource management. AI Agent Connect > OpenAI MCP gives your AI agent direct access to your OpenAI account. Instead of jumping back and forth to the web dashboard, you can manage fine-tuning jobs, track batch processing, inspect assistant configurations, and handle uploaded files through natural conversation in Claude, Cursor, or Windsurf. ## Overview - **Category:** ai-frontier - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_nUi0FAK4SIEESzN3EN6rUcUtM1G2aBwB1LndkjlH/ai-agent-connect - **Tags:** llm-management, fine-tuning, model-discovery, ai-assistants, file-management, api-orchestration ## Description You shouldn't have to leave your IDE or chat interface just to check if a fine-tuning job finished or to see why a batch failed. This MCP turns your AI client into a dedicated ML ops assistant. You can pull up a list of all your available models, check the status of active training runs, or manage the files you've uploaded for your Assistants. If a batch job is taking too long or a fine-tune went wrong, you can cancel them immediately without ever opening a browser. It puts your OpenAI infrastructure right where you actually work, making it easy to audit your assistants or clean up old files using simple text commands. ## Tools ### create_batch Starts a new batch processing job using a specific input file and endpoint. It returns a batch ID so you can track the progress. ### delete_file Removes an uploaded file from your OpenAI account. Use this carefully, as deleting a file can break any fine-tunes or assistants relying on it. ### get_assistant Retrieves the full configuration and details for a specific Assistant ID. ### get_batch Fetches the current status and details for a specific batch job. ### get_model Checks the metadata, ownership, and permissions for a specific model ID. ### list_assistants Shows all your configured Assistants, including their instructions, models, and enabled tools. ### list_batches Provides a list of all batch processing jobs and their current states, such as in_progress or completed. ### list_models Lists every model available to your account, including DALL-E, Whisper, and fine-tuned versions. ### get_fine_tune Returns the specific details and current status of a fine-tuning job. ### list_files Displays all files uploaded to your account, categorized by their purpose like fine-tuning or assistants. ### list_fine_tunes Shows all fine-tuning jobs and their current progress, base models, and estimated completion times. ### cancel_batch Stops a running batch job immediately. ### cancel_fine_tune Aborts a fine-tuning job if you need to stop training or if you used the wrong data. ## Prompt Examples **Prompt:** ``` Show me all available GPT models. ``` **Response:** ``` I found 12 GPT models in your account: gpt-4o, gpt-4o-mini, gpt-4-turbo, gpt-4, and gpt-3.5-turbo, plus several fine-tuned variants. Do you need details on a specific one? ``` **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 is training on 1,240 examples using gpt-3.5-turbo. It should finish in about 2 hours. ``` **Prompt:** ``` List all my uploaded files and their purposes. ``` **Response:** ``` You have 8 files uploaded: 3 for fine-tuning, 2 for Assistants, 2 for batch processing, and 1 vector store file. ``` ## Capabilities ### Model Discovery Your agent checks which models are available and what they can do. ### Job Monitoring Your agent tracks the real-time status of fine-tuning and batch processing. ### File Management Your agent browses and deletes files used for training or assistants. ### Assistant Auditing Your agent inspects the instructions and tools configured for your assistants. ### Batch Control Your agent creates or cancels bulk API processing jobs. ## Use Cases ### Fine-tuning Oversight Check if a training job succeeded or failed directly from your chat interface. ### Cost Management Run large scale requests using batch processing to keep API costs down. ### Assistant Configuration Verify the instructions and tools assigned to a specific assistant. ### Resource Cleanup Identify and delete old files used for previous fine-tuning runs. ## Benefits - Eliminates the need to switch to the OpenAI web dashboard. - Provides direct control over training jobs via natural language. - Centralizes model and file management within your existing workflow. - Allows for immediate cancellation of long-running or incorrect jobs. ## How It Works Connecting to your OpenAI account is a quick, one-time setup. 1. Subscribe to the OpenAI MCP on Vinkius. 2. Provide your OpenAI API Key. 3. Connect your preferred client like Claude or Cursor. 4. Start asking your agent to manage your models, files, and jobs. ## Frequently Asked Questions **How do I use this MCP with Claude?** Once you subscribe on Vinkius and enter your API key, you can connect it to Claude by following the standard MCP connection steps in your client. **Can I cancel a fine-tuning job?** Yes, you can use the cancel_fine_tune tool to stop a running job if you need to abort the process. **Is it safe to delete files through this MCP?** Deleting a file is irreversible. If you delete a file being used by an assistant or a fine-tune, those resources will break. **What models can I manage?** You can list and inspect any models available to your account, including GPT-4, DALL-E, Whisper, and your own fine-tuned models. **Do I need to use the OpenAI dashboard?** No, this MCP is designed so you can perform most management tasks through your AI agent instead of the web dashboard. **How do I get my OpenAI API Key?** Log in to the [**OpenAI Platform**](https://platform.openai.com/api-keys), go to **API Keys** in the left sidebar, click **Create new secret key**, give it a name and copy the key immediately — it starts with `sk-proj-` and won't be shown again. **Can I monitor my fine-tuning jobs?** Yes! Use `list_fine_tunes` to see all fine-tuning jobs with their status (validating_files, queued, running, succeeded, failed, cancelled). Use `get_fine_tune` with a specific job ID for detailed info including training progress, estimated finish time and result model ID. You can also cancel running jobs with `cancel_fine_tune`. **Can I manage batch processing jobs?** Yes! Use `list_batches` to see all batch jobs, `create_batch` to submit new batches with an input file ID and endpoint, `get_batch` to check progress and `cancel_batch` to stop running jobs. Batches process requests asynchronously at a lower cost than individual API calls. **Can I list and inspect my Assistants?** Yes! Use `list_assistants` to see all configured Assistants with their models, tools (code interpreter, file search, function calling) and instructions. Use `get_assistant` with a specific assistant ID for full details including file IDs and metadata.