# Dataiku DSS MCP for AI Agents AI Agent Connect > Dataiku DSS MCP. Manage your entire data science lifecycle, from project exploration and dataset schema checks to pipeline monitoring and model deployment, directly through your AI agent. It connects your AI client to your Dataiku instance so you can run automation scenarios and audit configurations using natural language. ## Overview - **Category:** ai-frontier - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_omB2MsN4Y1V0ZX0dD6ibopTA1n2aAb9aFGAtDQOY/ai-agent-connect - **Tags:** data-science, ml-ops, pipeline-orchestration, predictive-modeling, data-pipelines, automation ## Description This Connector gives your AI agent a direct line into your Dataiku DSS instance. Instead of jumping between tabs to check why a pipeline stalled or trying to remember the exact schema of a production dataset, you can just ask your agent for the status. It handles the heavy lifting of navigating your projects, pulling dataset details, and inspecting recipe configurations. You can use it to audit data logic in Python or SQL recipes, monitor the performance of saved ML models, and trigger automation scenarios to rebuild pipelines or retrain models on the fly. It's a way to bring your enterprise data science workflows into a single conversation. By connecting this through the Vinkius catalog, you get a unified way to oversee your organizational data connections and plugins without manual searching. It makes the bridge between your high-level goals and the actual data engineering tasks much shorter. ## Tools ### get_recipe Get the specific configuration and settings for a recipe. This is perfect for auditing logic. ### list_jobs View all pipeline jobs, including build tasks and training runs. It helps you track active work. ### get_job Get the state, timing, and output of a specific job. Use this to troubleshoot failed runs. ### list_connections List all data connections like databases or cloud storage. It helps you audit your data access points. ### list_projects See all projects available to your API key. It helps you quickly find the right workspace. ### get_project Pull metadata, settings, and tags for a specific project. Use this to understand project context. ### list_datasets Get a list of all datasets within a project. This is the first step for data exploration. ### dataset_schema Retrieve columns and types for a dataset. It saves you from manually checking table headers. ### list_recipes See all data transformation recipes in a project. Use this to browse your available logic. ### list_scenarios List all automation scenarios in a project. It shows you what can be triggered automatically. ### run_scenario Trigger a scenario execution to rebuild pipelines or retrain models. This automates your manual tasks. ### list_models See all deployed or saved ML models in a project. This is great for model inventory. ### get_model Get metadata, the algorithm used, and performance metrics for a model. Use it to evaluate results. ### list_plugins See all installed DSS plugins. Use this to check your environment's capabilities. ## Prompt Examples **Prompt:** ``` What's the status of the latest training run in the Churn project? ``` **Response:** ``` I've checked the jobs for the 'Churn-Prediction-Q1' project. The most recent training run finished successfully 12 minutes ago. It took 45 minutes to complete and the output is ready in the project's results folder. Would you like me to pull the performance metrics for that model? ``` **Prompt:** ``` Can you show me the columns in the 'raw_logs' dataset? ``` **Response:** ``` Here is the schema for the 'raw_logs' dataset in the FRAUD project: | Column Name | Type | | :--- | :--- | | timestamp | date | | user_id | string | | ip_address | string | | action_type | string | There are 12 columns in total. Let me know if you need the full mapping. ``` **Prompt:** ``` Trigger the pipeline rebuild for the Sales project. ``` **Response:** ``` Triggering the 'REBUILD_PIPELINE' scenario in the 'SALES' project now. The system is starting the rebuild of the underlying datasets and recipes. I'll keep an eye on it and let you know as soon as it's finished. ``` ## Capabilities ### Explore projects and datasets See all your active projects and pull column schemas for any dataset instantly. ### Monitor pipeline status Track build tasks and training runs to see execution states and timing. ### Audit recipe configurations Retrieve the exact logic for Python, SQL, and Visual recipes for verification. ### Trigger automation scenarios Start pipeline rebuilds or model retraining with simple text commands. ### Inspect ML model metrics View performance data and algorithm details for all saved models. ### Audit data connections Enumerate all database, cloud, and API connections in your environment. ## Use Cases ### Quick schema verification A data scientist needs to know if the 'raw_logs' table has a timestamp column. They ask the agent for the `dataset_schema`, and it confirms the column types instantly. ### Production job debugging An MLOps engineer sees a failed job and needs to know why. They ask the agent to check `get_job` for the last build task to see the error state and timing. ### Inventory auditing A manager wants to see all active projects. They ask the agent to `list_projects`, and it returns a list of all production and sandbox environments. ### Automated pipeline rebuilds A data engineer needs to trigger a weekly rebuild. They tell the agent to `run_scenario` for 'REBUILD_PIPELINE' in the 'SALES' project. ## Benefits - Faster Troubleshooting: Use `get_job` to see exactly where a pipeline stalled without digging through logs. - Rapid Data Exploration: Quickly grab `dataset_schema` details to understand your data structure before writing any code. - Automated Retraining: Trigger `run_scenario` to start model retraining or pipeline rebuilds using simple text commands. - Audit Ready Logic: Use `get_recipe` to pull exact configuration structures for SQL or Python recipes to verify data logic. - Centralized Model Oversight: Keep track of your entire model portfolio with `list_models` and `get_model` to see performance metrics at a glance. - Easier Connection Management: Quickly see all your `list_connections` to verify which databases and cloud storage sources are active. ## How It Works The bottom line is you get a natural language interface for your entire Dataiku environment. 1. Subscribe to this Connector on the Vinkius marketplace. 2. Provide your Dataiku Instance URL and API Key in your client settings. 3. Ask your agent to list projects, check job states, or run scenarios. ## Frequently Asked Questions **Can I use the Dataiku DSS MCP to see all my projects?** Yes, you can ask your agent to list every project your API key has access to. This helps you quickly navigate your workspace without manual searching. **How does the Dataiku DSS MCP help with MLOps?** It lets you monitor model performance metrics and trigger automation scenarios like retraining. You can manage the lifecycle of your models directly from your AI client. **Can I check the logic of a Dataiku recipe using this?** Yes, you can retrieve the configuration for Python, SQL, or Visual recipes. This is great for auditing data logic or debugging production pipelines. **Does the Dataiku DSS MCP support different data connections?** You can use it to list all your data connections, including databases, cloud storage, and APIs. It provides a clear overview of your organizational data access. **Can I run automation scenarios with the Dataiku DSS MCP?** You can trigger specific scenarios to rebuild pipelines or retrain models. It turns your manual maintenance tasks into simple natural language commands. **Is the Dataiku DSS MCP good for checking dataset schemas?** It's perfect for that. You can get the column names and types for any dataset in a project instantly, which speeds up your data exploration phase. **Can my agent trigger a Dataiku automation scenario?** Yes. Use the 'run_scenario' tool. Provide the project key and the scenario ID. The agent will command the backend to orchestrate the absolute workflow rules, triggering a new execution run for your pipeline or model retraining. **How do I check the schema of a specific dataset via chat?** Provide the project key and dataset name to the 'dataset_schema' tool. Your agent will validate the API arrays structurally and return the dataset column names and types natively, helping you understand your data boundaries. **Can I monitor the performance of saved ML models?** Absolutely. Use the 'get_model' tool. Your agent retrieves the metadata and performance metrics defining specific trained schema layers, allowing you to audit model quality and drift without opening the DSS UI.