# Mistral AI MCP for AI Agents AI Agent Connect > Mistral AI provides a direct bridge to Mistral's suite of models, including Mistral Large, Small, and Codestral. This Connector lets your AI agent handle chat completions, generate vector embeddings, and run large-scale batch jobs. You can also use it to moderate content for safety or manage uploaded files for document processing. It's the quickest way to put Mistral's inference power into your existing AI workflow. ## Overview - **Category:** ai-frontier - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_DiLO9sALvzBLtPCvHvpEl9MvzgvDgoveKQ7OqBuA/ai-agent-connect - **Tags:** large-language-models, embeddings, natural-language-processing, generative-ai, api-integration, rag ## Description This Connector puts Mistral's entire model lineup into your AI agent's hands. You can have your agent switch between high-reasoning models for complex tasks or smaller, faster models for simple classification. It handles the heavy lifting of model discovery and batch processing so you don't have to manually manage job IDs or monitor status updates. When you're building something that needs to scale, you can fire off huge sets of prompts as batch jobs and let the system handle the queue. For teams needing to keep things clean, the content moderation tools let your agent check for safety issues before anything goes live. It's a core part of how I manage multiple model providers through the Vinkius catalog, making it easy to swap out providers without rewriting your entire logic. You get a unified way to generate embeddings for your search index and manage your files in one place. Instead of dealing with various API endpoints for different tasks, you just talk to your agent. It handles the logic of choosing the right model, checking the safety of the input, and managing the lifecycle of your batch jobs. This means you spend less time on the plumbing and more time on the actual features you're trying to build. Whether you're trying to summarize thousands of documents or just need a quick response from a specialized coding model, this connection makes it all accessible from a single interface. ## Tools ### chat_completion Generate text using Mistral models ### list_models See every Mistral model available to your account along with its context window. This helps you discover models. ### analyze_sentiment Analyze text sentiment ### summarize_text Summarize long documents ### translate_text Translate text between languages ### fix_grammar Correct grammar and spelling ### create_embeddings Generate vector embeddings ### explain_code Explain logic in code ### extract_entities Extract data as JSON ### generate_code Write code snippets ## Prompt Examples **Prompt:** ``` What models can I use right now? ``` **Response:** ``` Here are the Mistral models currently available to your account: | Model ID | Capabilities | Context Window | | :--- | :--- | :--- | | **mistral-large-latest** | Most capable, high reasoning | 128K | | **mistral-small-latest** | Efficient, fast inference | 32K | | **codestral-latest** | Code specialist | 32K | | **mistral-embed** | Vector embeddings | 8K | ``` **Prompt:** ``` Check if this user comment is safe to post: 'I really hate everyone and want to cause trouble.' ``` **Response:** ``` I've checked that text for safety. Here are the scores: * **Violence:** 0.98 * **Hate:** 0.95 * **Sexual:** 0.01 * **Self-harm:** 0.00 **Result:** This content is flagged as unsafe due to high violence and hate scores. ``` **Prompt:** ``` I have a file of 500 reviews. How do I process them? ``` **Response:** ``` For a large set like that, you should use the **batch processing** feature. Here's how we can do it: 1. Upload your JSONL file containing the 500 reviews. 2. I'll use `create_batch` to start the job. 3. I can then use `list_batches` to keep an eye on the progress for you. This is much more cost-effective than sending them one by one. ``` ## Capabilities ### Chat with Mistral models Send prompts to various models and get back responses with custom temperature or token limits. ### Run high-volume batch jobs Submit large sets of prompts for asynchronous processing to handle heavy workloads efficiently. ### Generate vector embeddings Turn text into numbers for your search engine or similarity comparison systems. ### Moderate user content Get safety scores for text to filter out harmful categories like violence or hate speech. ### Manage uploaded files List and delete the files you use for fine-tuning or batch processing in one place. ### Discover model capabilities See all available models along with their specific context windows and IDs. ## Use Cases ### Scaling product descriptions A developer needs to generate 10,000 product descriptions and uses create_batch to do it cheaply and efficiently. ### Automated safety filtering A content team wants to filter out hate speech and uses moderate to check every comment before it gets posted. ### Model capability auditing An ML engineer is testing different models and uses list_models to see which one has the best context window for their data. ### Building semantic search A search engineer needs to index a new blog and uses embeddings to create a vector database for the site. ## Benefits - Switch models instantly using chat to pick the right balance of speed and reasoning for every specific task. - Save money on high-volume tasks by using create_batch to process thousands of prompts asynchronously. - Build smarter search features by using embeddings to turn your data into searchable vectors. - Keep your platform safe by using moderate to automatically score user content for harmful categories. - Simplify your workflow by using list_models to see all available context windows and capabilities in one view. - Manage your data better by using list_files and delete_file to keep your Mistral storage clean. ## How It Works The bottom line is you get a direct line to Mistral's model family without managing separate API calls. 1. Connect your Mistral AI API key to the Connector. 2. Ask your AI agent to perform a task like chatting with a specific model or starting a batch. 3. The agent executes the request and returns the response or status directly in your chat. ## Frequently Asked Questions **Can I use the Mistral AI MCP with Cursor?** Yes, you can connect this Connector to Cursor or any other MCP-compatible client like Windsurf or VS Code to access Mistral's models directly. **Does the Mistral AI MCP support large-scale batch processing?** Yes, it allows you to submit large sets of prompts as batch jobs, which is ideal for high-volume tasks like data extraction or summarization. **How do I get my Mistral AI API key for the Connector?** You can get your API key directly from the Mistral AI dashboard. Once you have it, just paste it into your Connector configuration. **Can I use Mistral AI MCP for generating vector embeddings?** Yes, the Connector includes a dedicated tool to convert text into vector embeddings, which you can use for building semantic search or similarity checks. **Is the Mistral AI MCP good for content moderation?** It's great for that. You can have your agent automatically score text for categories like hate speech, violence, and self-harm to keep your platform safe. **How do I manage my uploaded files with the Mistral AI MCP?** You can use the Connector to list all your uploaded files and delete specific ones you no longer need, helping you keep your Mistral storage organized. **Can I switch between different Mistral models using this Connector?** Yes, you can easily switch between models like Mistral Large for complex logic and Codestral for coding tasks by just asking your agent. **How do I get a Mistral AI API Key?** Log in to the [**Mistral Console**](https://console.mistral.ai), go to **API Keys** in your workspace settings, click **Create new key** and copy it immediately. You'll need to set up billing in the admin portal first. **What models are available?** Use the `list_models` tool to see all available Mistral models. Key models include mistral-large-latest (most capable), mistral-small-latest (efficient), codestral-latest (code specialist), and mistral-embed for embeddings. Each has different context windows, capabilities and pricing. **Can I send multi-turn conversations?** Yes! Pass a messages array with alternating 'user', 'assistant' and 'system' roles. Each message has a 'role' and 'content' field. Mistral will continue the conversation based on the full message history. **Can I moderate content for safety?** Yes! Use the `moderate` tool with text input. It returns safety scores for categories including sexual, hate, violence, self-harm, criminal and other harmful content. This is useful for filtering user-generated content before processing. **Which models can I access?** Access all available endpoints including `mistral-large-latest`, `mistral-small-latest`, `open-mixtral-8x22b`, and `mistral-embed`. **How does Mistral authentication work?** Mistral requires an **API Key** sent as a Bearer token against `api.mistral.ai/v1`. **Can I generate vector embeddings?** Yes. Use the `mistral-embed` model to generate 1024-dimensional embeddings for your text data.