# Gradient AI MCP for AI Agents AI Agent Connect > Gradient AI MCP lets you manage fine-tuned models, run advanced completions, and handle NLP tasks like sentiment analysis or entity extraction. It gives your AI agent direct access to Gradient AI's infrastructure for things like RAG collections, audio transcriptions, and high-dimensional embeddings. ## Overview - **Category:** developer-tools - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_KwJfHerXBfsWKXVQH8bGJG4HGUwUfVlRXkPpGPNU/ai-agent-connect - **Tags:** llm, fine-tuning, embeddings, nlp, ai-infrastructure ## Description This Connector lets you manage fine-tuned models and high-quality completions using Gradient AI's infrastructure. You can move past the headache of manual data processing by using tools for sentiment analysis and entity extraction directly from your AI client. It handles the heavy lifting of training models on your own datasets to improve performance on niche tasks. You can also convert text into high-dimensional vectors for search or set up RAG collections to query your documents. It's one of the many reliable connections in the Vinkius catalog that lets you move from a prototype to a production-ready system without the friction of managing complex backend logic. ## Tools ### analyze_sentiment Tells you the emotional tone of a document. It's useful for categorizing customer feedback at scale. ### answer_question Pulls specific answers from a source document. Use this for building Q&A bots on your internal docs. ### complete_model Generates text based on a prompt. It handles the core task of getting a response from your chosen model. ### fine_tune_model Trains a model on a specific set of samples. Use this to teach the model niche behaviors or styles. ### generate_embeddings Turns text into high-dimensional vectors. This is the standard way to prepare data for similarity search. ### get_model Shows the details of a specific model. Use this to check parameters or status during an experiment. ### get_transcription Checks the results of a transcription job. Use this to see if your audio has been successfully processed. ### list_embeddings Shows which models you can use for embeddings. This helps you pick the right vector size for your project. ### list_models Shows all foundational and fine-tuned models. It's the quickest way to see what's currently available to you. ### list_rag_collections Shows all RAG collections in your workspace. Use this to manage multiple knowledge bases at once. ### personalize_document Rewrites a document for a specific audience. It helps tailor content for different user segments. ### summarize_document Creates a short summary of a long document. It saves time when you need to digest large amounts of text. ### upload_file Sends a file to the cloud for processing. It's the first step for any operation involving PDFs or large docs. ### create_model Sets up a new fine-tuned model instance. Use this to start a new training run with specific parameters. ### create_rag_collection Builds a collection for RAG operations. It organizes your data so your agent can find relevant info quickly. ### create_transcription Starts a job to turn audio into text. It handles the heavy lifting of speech-to-text processing. ### delete_model Removes a fine-tuned model from your workspace. Use this to keep your environment clean of old experiments. ### extract_entity Pulls structured data out of a document based on a schema. It turns messy text into clean, usable fields. ### extract_pdf Pulls text and data out of PDF files. It's the primary way to ingest static documents into your system. ## Prompt Examples **Prompt:** ``` I need to see what models I have available in my workspace. ``` **Response:** ``` I've checked your Gradient AI workspace. Here's what you have access to: | Model Type | Model Name | Status | | :--- | :--- | :--- | | Foundational | llama3-8b | Active | | Fine-tuned | support-v1-final | Ready | You can use either of these for your next task. ``` **Prompt:** ``` Can you summarize this text and tell me the sentiment? ``` **Response:** ``` Here is the summary of the text you provided: **Summary:** The user is reporting a bug in the login flow, specifically mentioning a timeout error on the mobile app. **Sentiment:** Negative. The user expressed frustration with the current experience. ``` **Prompt:** ``` Extract the names and dates from this PDF. ``` **Response:** ``` I've processed the document for you. Here is the structured data I found: * **Name:** John Doe * **Date:** October 12, 2023 * **Reference Number:** #99210 I've extracted these into a clean format for your records. ``` ## Capabilities ### Train custom models Fine-tune models on your specific datasets to handle niche tasks or brand-specific voices. ### Convert text to vectors Generate high-dimensional embeddings for use in similarity searches and vector databases. ### Extract structured data Turn messy documents into clean, usable fields using entity extraction. ### Automate audio transcription Start and monitor jobs to turn audio files into searchable text automatically. ### Manage RAG collections Build and organize collections for retrieval augmented generation tasks. ### Analyze document sentiment Quickly determine the emotional tone of text for feedback analysis at scale. ## Use Cases ### Building a customer support bot A dev needs to answer questions from a 200-page manual. They use extract_pdf to ingest the data and create_rag_collection to make it searchable. ### Sentiment analysis at scale A marketing lead wants to know how people feel about a product launch. They use analyze_sentiment to scan thousands of reviews in seconds. ### Training a niche model An engineer needs a model that speaks in a very specific brand voice. They use fine_tune_model to train on historical data and then list_models to verify the result. ### Automated transcription A content creator wants to turn hours of interviews into text. They use create_transcription to process the audio and get_transcription to pull the final text. ## Benefits - Speed up model deployment by using create_model and fine_tune_model to automate your training pipeline. - Build production RAG systems faster by managing collections with create_rag_collection and list_rag_collections. - Save hours of manual data cleaning by using extract_entity to turn messy documents into structured data. - Automate content processing with summarize_document and personalize_document to handle large volumes of text. - Simplify audio workflows by using create_transcription to handle speech-to-text tasks without managing local libraries. - Get high-quality vectors for your search engine by using generate_embeddings and list_embeddings. ## How It Works The bottom line is you get a direct pipeline to Gradient AI's infrastructure without leaving your AI client. 1. Connect your Gradient API Key and Workspace ID in your AI client. 2. Tell your agent to list available foundational models or upload your data files. 3. Start a fine-tuning job, generate embeddings, or create a RAG collection. ## Frequently Asked Questions **Can I use the Gradient AI MCP to train models on my own data?** Yes, you can use the fine_tune_model tool to train models on your specific datasets. This allows your agent to handle the training process and manage the resulting models directly within your workspace. **How does the Gradient AI MCP handle PDF files?** The Connector includes a tool to extract text and data from PDFs. Your agent can pull the content out so you can then summarize it, ask questions about it, or extract specific entities. **Can I use this Connector to build a RAG system?** Absolutely. You can create and manage RAG collections using the Connector. This makes it easy to set up a knowledge base for your agent to query during conversations. **Does the Gradient AI MCP support audio transcription?** Yes, it can start transcription jobs and retrieve the results. This is great for turning meetings or interviews into searchable text automatically. **How do I manage my fine-tuned models with the Gradient AI MCP?** You can list all your models to see what's available, get specific details on any model, or delete old versions. It gives you full control over your model library. **Can the Gradient AI MCP do sentiment analysis?** Yes, it has a dedicated tool for analyzing the sentiment of documents. Your agent can use this to quickly gauge the tone of customer feedback or reviews. **How can I start training a custom model with my own data?** You can use the `fine_tune_model` tool. Simply provide the model ID and an array of training samples. The agent will handle the submission to Gradient's training infrastructure. **Can I use RAG (Retrieval Augmented Generation) with this server?** Yes! The `complete_model` tool includes an optional `rag` parameter, allowing you to provide context or collection IDs to ground the model's responses in specific data. **How do I generate vector embeddings for my documents?** Use the `generate_embeddings` tool by specifying a model slug (like 'bge-large') and a list of text inputs. It will return the high-dimensional vectors for your text.