# DVC MCP for AI Agents AI Agent Connect > DVC lets you manage machine learning experiments and data versioning directly through your AI agent. Track model runs, audit metric histories, and navigate organization workspaces in DVC Studio without switching tabs. It connects your DVC Studio account to your workflow, allowing you to pull metric arrays, audit experiment histories, and navigate organization workspaces using natural language. ## Overview - **Category:** developer-tools - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_Myft115Qe8hEZoVFuhiTYRKEW5euUlsOR2XqjK69/ai-agent-connect - **Tags:** machine-learning, version-control, experiment-tracking, data-pipelines, model-management, git-workflow ## Description Managing machine learning models usually means jumping between dashboards, terminal logs, and spreadsheets to see which experiment actually worked. This Connector changes that by bringing your DVC Studio data directly into your chat interface. You can ask your agent to pull specific metric arrays from past runs or check which repositories are connected to your workspace. It handles the heavy lifting of navigating complex project structures so you can stay focused on the actual science. Because Vinkius hosts this alongside thousands of other tools, you can keep your entire dev stack connected in one spot. Instead of hunting for a specific run ID, you just ask for the best performing model from last week and get the data you need immediately. You get a clear picture of your ML footprint and can verify permissions or project scopes without leaving your chat interface. ## Tools ### list_views Show all active dashboard layouts in your workspace. This helps you see how your data is structured. ### get_view Pull the configuration for a specific dashboard. Use this to see the exact layout of a view. ### list_projects List every project in your organization. Use this to get a high-level overview of your work. ### get_project Fetch the metadata for a specific project. It gives you the details you need for a single repository. ### list_experiments Show all model runs and experiments. This lets you see the history of your training cycles. ### get_user Retrieve the profile of a specific user. This is useful for checking identity and permissions. ## Prompt Examples **Prompt:** ``` What was the best accuracy for the Image-Classification-V2 project? ``` **Response:** ``` I've checked your latest runs. The best result was **0.96 accuracy** from the experiment finished yesterday. Would you like to see the full metrics for that run? ``` **Prompt:** ``` Who has admin roles in our organization? ``` **Response:** ``` I found the following users with administrative permissions: - **Jane Doe** (Lead ML Engineer) - **John Smith** (DevOps Lead) - **Sarah Chen** (Data Science Manager) ``` **Prompt:** ``` List all the dashboard views available to me. ``` **Response:** ``` You have 3 active dashboard views: - **Training_Metrics_Main** - **Validation_Plots_V2** - **Data_Drift_Summary** Which one would you like to see the configuration for? ``` ## Capabilities ### Browse active dashboard layouts See how your team visualizes data without opening the UI. ### Pull specific project details Get the metadata for any repository in your workspace. ### View user profile permissions Identify who has access to your sensitive ML datasets. ### Browse organization projects Get a high-level overview of every project in your workspace. ### Fetch individual project data Retrieve the specific constraints and details of a single repo. ### Scan previous experiments See the full history of your training cycles and model runs. ## Use Cases ### Comparing model accuracy A data scientist asks for the best accuracy from the last 5 runs to decide which model to deploy. ### Onboarding a new team member An ML engineer checks what permissions the new hire has to ensure they can access the right datasets. ### Auditing project history A team lead asks for a list of all projects to see what is currently in the production pipeline. ### Debugging a broken pipeline A DevOps engineer checks if the repository is actually connected to verify a failed integration. ## Benefits - Stop hunting for specific model runs by seeing every training cycle in a single list. This lets you quickly identify the best performers without digging through old logs. - Get a high-level overview of your entire organization's work by browsing every repository in your workspace. You can see all projects instantly to ensure your agent is looking at the right data. - Verify team permissions and identify authorized token holders to manage who has access to sensitive machine learning datasets. This removes the guesswork when checking project scopes. - Pull specific dashboard layouts to understand how your team visualizes key metrics. It gives you the structural representation of your workspace without having to open the UI. - Retrieve complex metric arrays from past runs to compare performance across different training epochs. This lets you do deep data analysis through your AI client instead of manual copy-pasting. ## How It Works The bottom line is you get a conversational interface for your entire DVC Studio experiment history. 1. Subscribe to the DVC MCP on the Vinkius marketplace. 2. Copy your Client Access Token from your DVC Studio Profile Settings. 3. Paste the token into your AI client to start querying your ML data. ## Frequently Asked Questions **How can I use DVC MCP to track my ML models?** You can use it to query your experiment history and metrics directly through your AI client. It connects to DVC Studio so you can ask questions about your past runs and model performance in natural language. **Can I see my experiment history with DVC MCP?** Yes, you can list all previous experiments and view specific metric arrays. This helps you compare different training cycles without leaving your chat interface. **Does DVC MCP work with my current DVC Studio account?** It does. You just need to provide your DVC Studio Client Access Token. Once connected, your AI agent can access your projects, views, and experiment data. **How do I check my project permissions using DVC MCP?** The Connector allows you to retrieve user profiles and check identity roles. You can ask your agent who has access to specific projects or what permissions a team member holds. **Can I get specific metrics from past runs?** Yes, you can retrieve complex metric arrays from specific experiment epochs. This is great for doing deep data analysis on your best performing models. **Is DVC MCP good for team leads?** It's excellent for team leads who need to monitor organization workspaces. You can quickly list all projects and check the progress of multiple experiments across the team. **Can my agent list all experiments for a specific DVC project?** Yes. Use the 'list_experiments' tool. Provide the project ID, and the agent will iterate through the model runs, returning a detailed history of metrics and execution logs for that project. **How do I see my custom dashboard views via chat?** Use the 'list_views' tool to see all your dashboard layouts. You can then use 'get_view' with a specific ID to retrieve the structural configuration and settings for that exact UI representation. **Can I audit my DVC Studio project settings through the agent?** Absolutely. The 'list_projects' and 'get_project' tools allow your agent to analyze identifier boundaries and repository metadata, helping you verify project connections and team mappings natively.