OpenAI Alternative MCP. Manage models and pipelines without leaving your chat client.

You're not messing with a basic API wrapper here; this setup lets your agent handle every complex resource operation you need for OpenAI models, keeping everything visible in one place. You can manage model lifecycles, pipeline jobs, and configured assistants without ever having to jump to the native dashboard.

Checking Models and Metadata
To audit what's available, use list_models—it spits out every single model ID your account has access to, like GPT-4o or text-embedding. Before you run anything, you can check the metadata for a specific model using get_model; this gives you the ID, owner, and creation date so you know exactly what you're working with.

Managing Assistants
You need to see what agents you’ve built? Run list_assistants to get an overview of every configured Assistant; that list shows you their IDs, instructions, base models, and which tools they use. If you want the full rundown on just one agent, use get_assistant with its ID to retrieve all the detailed metadata.

Handling Data Batches
Need to run heavy data processing? You start a new job by calling create_batch, where you specify the input file and the API endpoint. To track what's happening, check the status of an existing job using get_batch; this gives you detailed metrics and status updates for a known batch ID.

If that batch goes sideways or runs too long, don't worry—you can stop it dead in its tracks by running cancel_batch with the corresponding batch ID.

Controlling Fine-Tuning Pipelines
Model training requires careful tracking. To get a list of all your fine-tuning jobs, use list_fine_tunes; this shows you IDs, base models, current status, and even an estimated finish time for every job running. If you need to know the deep details on one specific run, call get_fine_tune with its ID.

And if a job gets stuck or you realize you uploaded the wrong training file, use cancel_fine_tune to halt the process immediately.

Tracking and Cleaning Up Files
You've got files for fine-tuning, batch jobs, and Assistants—you need to keep track of 'em. Use list_files to see every asset uploaded to OpenAI; this list gives you the ID, purpose (like whether it’s for a fine-tune or batch), size, and current status. When you're done with an asset and want it gone forever, run delete_file; remember that action is permanent and breaks any related Assistants or fine-tunes.

The Bottom Line
Your agent handles all this complexity for you. You just tell your client what you need—like 'list my models' or 'stop the batch job ID 123'—and it runs the multi-step API calls needed to get the job done.