# IBM watsonx MCP for AI Agents AI Agent Connect > IBM watsonx MCP connects your AI agent to IBM's enterprise foundation models. It lets you manage projects, list models, generate text, and start tuning jobs directly from your chat interface. Use it to bridge the gap between your internal data and IBM's production-ready AI infrastructure. ## Overview - **Category:** industry-titans - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_gXytFT8fNy6ELfNWb5ZhGBzw4TcaHPZqjIMO9o5b/ai-agent-connect ## Description This Connector puts the full power of IBM watsonx into your hands without making you jump between different browser tabs and dashboards. Instead of manually logging into a console to check model specs or start a tuning job, you can just ask your agent to do it for you. You can pull in specific foundation models for your project, check on the progress of your prompt tuning, or generate embeddings for your own search index. It's built for teams who need to keep their AI workflows grounded in enterprise-grade infrastructure. By adding this to your setup via the Vinkius catalog, you turn your AI client into a control center for your entire watsonx environment. You can create prompts, generate multi-turn chat completions, and summarize large blocks of text without ever leaving your workspace. It bridges the gap between your internal data and production-ready AI infrastructure, making it easy to manage models at scale. It handles the heavy lifting of model lifecycle management. You can see what's available, get the technical details of a specific model, and start tuning your prompts with just a few words. Whether you're building a custom chatbot or a complex data analysis pipeline, this connection ensures you're always using the right tools for the job. It removes the friction of navigating complex cloud consoles, letting you stay focused on the actual work of building and deploying. It's about having a direct line to your enterprise assets without the overhead. ## Tools ### create_prompt Create a new prompt in your watsonx project. It saves your instructions for later use. ### generate_chat Generate multi-turn chat completions using a watsonx chat model. This is perfect for building conversational apps. ### generate_embeddings Turn your text into vector embeddings. Use these for similarity searches and clustering your data. ### generate_text Create text for summaries or content generation. It uses a foundation model to produce single-turn outputs. ### get_model_details See the full specs for a specific foundation model. It shows you the details you need to make a choice. ### get_tuning_status Check if your prompt tuning job is finished. It returns the current status of your active job. ### list_models See all the foundation models available in watsonx. It shows you IDs, families, and capabilities. ### list_projects Get a list of all your projects. This helps you keep track of your different workspaces. ### list_prompts See every prompt you've saved in a specific project. It makes it easy to find your best instructions. ### start_model_tuning Start a prompt tuning job for a model. You just need to provide a URL to your training data in cloud storage. ## Prompt Examples **Prompt:** ``` What models can I use for a summarization task? ``` **Response:** ``` Here are the top models in your watsonx environment suitable for summarization: | Model Family | Capability | Status | | :--- | :--- | :--- | | **LLM-Large** | High-quality summarization | Active | | **LLM-Medium** | Fast summarization | Active | Would you like me to get the full technical specs for one of these? ``` **Prompt:** ``` Check the status of my tuning jobs. ``` **Response:** ``` I've checked the current status of your active jobs: * **Job ID: 88291** - Status: *In Progress* (65% complete) * **Job ID: 99302** - Status: *Completed* Let me know if you want me to start a new one or get more details on a specific ID. ``` **Prompt:** ``` Create a prompt for a customer service bot. ``` **Response:** ``` I've created a new prompt for your customer service bot in your current project. **Prompt Name:** `CS_Bot_v1` **Description:** Handles basic inquiries and routes complex issues to humans. It's now saved in your project and ready for testing. ``` ## Capabilities ### List available foundation models See all available models in your watsonx environment. ### Generate text for summarization Create summaries or long-form content using foundation models. ### Start prompt tuning jobs Kick off model tuning using data stored in your cloud storage. ### Create and manage prompts Save and organize your best prompts within specific projects. ### Generate vector embeddings Turn text into vectors for similarity search and clustering. ### Check tuning progress Get real-time status updates on your active tuning jobs. ### Get model specifications Pull technical details for any specific model in the catalog. ## Use Cases ### Checking model specs for a new project An ML engineer needs to know which models fit their requirements. They ask their agent to list models and get the technical details of the best options. ### Summarizing large batches of content A content creator has a pile of transcripts. They ask their agent to summarize the text using a foundation model to get the key points instantly. ### Monitoring training jobs A developer starts a prompt tuning job and wants to know when it's done. They ask their agent to check the status while they continue coding. ### Building a semantic search index A researcher needs to build a RAG system. They ask their agent to turn their documentation into vector embeddings for a similarity search. ## Benefits - Stop switching tabs by seeing what's available directly in your chat. - Speed up research by checking model specs without leaving your IDE. - Automate content creation with instant summaries or drafts. - Manage your library to organize your best instructions. - Monitor your training progress while you focus on other tasks. - Build better search tools to handle semantic analysis. ## How It Works The bottom line is you get a direct bridge to IBM's enterprise AI models inside your favorite workspace. 1. Connect your IBM watsonx credentials to your AI client. 2. Tell your agent which project or model you want to work with. 3. Execute tasks like generating text or starting a tuning job. ## Frequently Asked Questions **Can I use IBM watsonx with my AI agent?** Yes, this Connector connects your AI agent directly to IBM watsonx, letting you manage models and prompts from your preferred chat interface. **How do I start a prompt tuning job using this Connector?** Just tell your agent to start a tuning job and provide the URL for your training data in cloud storage. It handles the rest. **Can this Connector generate vector embeddings?** Yes, you can use it to turn text into vector embeddings for use in similarity searches, clustering, and other semantic analysis tasks. **Is this for production AI workflows?** Exactly. It's designed for enterprise environments that need to manage foundation models and prompt tuning at scale. **Can I see which models are available in my account?** Yes, you can ask your agent to list the foundation models in your watsonx environment to see their IDs, families, and capabilities. **Does this work with Claude or Cursor?** Yes, it works with any MCP-compatible client, including Claude, Cursor, Windsurf, and VS Code.