DVC MCP. Track ML projects and audit model run history.
Gemini Works with every AI agent you already use
…and any MCP-compatible client
Just plug in your AI agents and start using Vinkius.
DVC: Manage ML experiments and data versions using your AI client. Connect your DVC Studio account to track projects, audit model runs, and inspect metrics without leaving your agent.
It lets you list projects, check experiment histories, and pull structural view configurations—all through natural conversation.
What your AI agents can do
Get project
Retrieves metadata for a specific ML project within DVC Studio.
Get user
Fetches the profile details for a given user account.
Get view
Retrieves the structural configuration and settings for a specific dashboard view.
Retrieves a list of all registered projects within the DVC Studio account.
Retrieves information about a single, defined project.
Retrieves a list of defined dashboard views and their settings.
Retrieves a list of past model runs and experiments in a given project.
Retrieves the profile information for a specified user.
Retrieves the detailed structural settings for a specific dashboard view.
Ask AI about this MCP
Supported MCP Clients
Gemini Waiting for input…
DVC MCP Server: 6 Tools for ML Experiment Tracking
These tools let your AI agent read, list, and retrieve structured data about ML projects, experiments, and dashboard views in DVC Studio.
019d758aget project
Retrieves metadata for a specific ML project within DVC Studio.
019d758aget user
Fetches the profile details for a given user account.
019d758aget view
Retrieves the structural configuration and settings for a specific dashboard view.
019d758alist experiments
Lists all past model runs, including metrics and run IDs, for a specific project.
019d758alist projects
Provides a list of all registered projects in the DVC Studio account.
019d758alist views
Lists all defined dashboard views available in the workspace.
Choose How to Get Started
Build a custom MCP for your own tools, or connect a ready-made integration from our catalog.
Build Your Own
Turn any API into an MCP. Import a spec, define Agent Skills, or deploy with MCPFusion.
- Import from OpenAPI, Swagger, or YAML specs
- Create Agent Skills with progressive disclosure
- Deploy to edge with MCPFusion framework
- Built in DLP, auth, and compliance on every call
- Real time usage dashboard and cost metering
- Publish to catalog or keep private
Make Your AI Do More
Start with DVC, then connect any of our 4,700+ other servers whenever your AI needs more. One click, no limits.
- Use this MCP plus 4,700+ others, all in one place
- Add new capabilities to your AI anytime you want
- Every connection is secured and compliant automatically
- Track usage and costs across all your servers
- Works with Claude, ChatGPT, Cursor, and more
- New servers added to the catalog every week
What you can do with this MCP connector
You'll connect your DVC Studio account to your AI client. This server gives your agent tools to track your ML projects and experiments. You can use it to list all projects registered in DVC Studio. You can also get metadata for one specific ML project. If you need to know who's using it, your agent can fetch a user's profile details.
To see what dashboards you've set up, your agent lists all defined dashboard views. You can get the structural configuration and settings for a specific dashboard view. For model runs, your agent lists all past experiments, pulling metrics and run IDs for a project. You can also look up the structural view configurations by getting a specific view's settings.
How DVC MCP Works
- 1 Subscribe to the DVC server and provide your DVC Studio Client Access Token.
- 2 Ask your AI agent to perform a task, like 'List all projects' or 'Show recent experiments for X'.
- 3 The agent uses the appropriate tool, and you receive structured data detailing the project, runs, or metrics.
The bottom line is, you manage your entire ML data lifecycle by talking to your agent, which handles the structured calls to DVC Studio for you.
Who Is DVC MCP For?
Data Scientists, ML Engineers, and Team Leads who need to track, audit, and compare model runs without leaving their development environment. This is for the data professional who's tired of clicking through multiple dashboards and version control systems just to check a metric.
Monitors model experiments and tracks performance metrics across different runs, pulling history without leaving the agent chat.
Verifies project repository connections and audits model run histories using natural language prompts to check for specific data constraints.
Gets a high-level overview of organization workspaces and experiment progress by asking conversational questions.
What Changes When You Connect
- See the complete list of your ML projects using
list_projects. You can immediately check if a new workspace is registered and start auditing its history. - Audit model runs easily. Running
list_experimentslets you see the last five runs and compare key metrics (like accuracy) across different epochs without diving into a GUI. - Understand your data structure by running
list_views. This tells you exactly what dashboard layouts exist and what kind of metrics they track. - Verify user permissions and team scope using
get_user. You can confirm who holds the access token and what roles they have before running complex operations. - Deep dive into specific data points using
get_project. You can pull project metadata to understand the full scope of the models being tracked. - Get granular details on dashboard settings by calling
get_view. This gives you the structural representation of a view, which is critical for debugging data pipelines.
Real-World Use Cases
Checking model performance across versions
A data scientist needs to know which model run was best. They prompt their agent: 'Show me the last 5 experiments for the Credit-Scoring-Model.' The agent calls list_experiments, retrieving key metrics and identifying the best-performing run ID immediately.
Onboarding a new team member
An ML engineer needs to verify the new hire has access. They ask the agent to run get_user and then get_project for the new team member's account. The agent verifies the user's scope and checks the project's existence before allowing further work.
Debugging a broken dashboard
A team lead spots a dashboard showing incorrect data. They prompt the agent to run get_view and get_view for the project. The agent extracts the detailed UI configuration, allowing the lead to pinpoint which structural setting broke the data feed.
Auditing repository connections
A DevOps engineer needs to confirm all required repositories are connected. They use list_projects to see all registered workspaces, then use get_project to validate the physical connection status for each one.
The Tradeoffs
Manual DVC UI navigation
Trying to check run history by clicking through the DVC Studio GUI, opening the project dashboard, and manually scrolling through metrics to find the best accuracy score.
→
Just ask your agent. Prompt it: 'List the last 5 experiments for [Project Name].' The agent runs list_experiments and gives you the metrics directly in the chat.
Relying on tribal knowledge
Asking a colleague, 'Do you know where the settings for the validation dashboard are?' and spending 15 minutes navigating the internal documentation to find the right structural IDs.
→
Use the agent to run list_views, then run get_view on the specific view name. You get the exact, structured settings immediately.
Sequential API polling
Writing a script that first calls the project API, then the user API, then the view API, and then tries to correlate the data manually in Python, risking key mismatches.
→
Let your agent handle the sequence. If you need project scope and user roles, ask for it in one prompt. The agent uses get_project and get_user and returns a correlated summary.
When It Fits, When It Doesn't
Use this server if your core workflow involves tracking, comparing, or auditing machine learning experiments and their associated metadata. You need to answer questions like: 'What were the metrics for model X's run in Q3?' or 'Are all our required projects connected?'
Don't use this if your task is simple data entry or live data transformation. If you only need to check if a project exists, list_projects is enough. If you need to know the specific configuration of a dashboard, use get_view. If you're debugging a specific run's performance, list_experiments is your starting point. If you need to know who has permission, run get_user. The tools work together to give you a full picture of the ML lifecycle, but you must target the specific tool for the specific piece of information you need.
Independent Platform Disclaimer: Vinkius is an independent platform and is not affiliated with, endorsed by, sponsored by, verified by, or otherwise authorized by DVC. All third-party trademarks, logos, and brand names are the property of their respective owners. Their use on this website is strictly for informational purposes to identify service compatibility and interoperability.
VINKIUS INFRASTRUCTURE
Cloud Hosted
Managed infra
V8 Isolated
Sandboxed per request
Zero-Trust Proxy
No stored credentials
DLP Enforced
Policy on every call
GDPR Compliant
EU data residency
Token Compression
~60% cost reduction
Works with Claude, ChatGPT, Cursor, and more
The Model Context Protocol standardizes how applications expose capabilities to LLMs. Instead of operating in isolation, your AI gains direct access to external platforms, live data, and real-world actions through secure, standardized connections.
This server provides 6 capabilities that interface natively with Claude, ChatGPT, Cursor, and any MCP client. No middleware. No custom integration required.
Available Capabilities
Sifting through DVC Studio's project metadata is a nightmare.
Today, figuring out the scope of an ML project means logging into DVC Studio. You click the project, check the dashboard, then manually navigate to the version history to see what metrics were logged. You have to open multiple tabs just to piece together whether the model run was valid and which metrics were captured.
With this MCP Server, you just ask your agent: 'Show me the project metadata and the last 5 experiments for [Project Name].' The agent uses `get_project` and `list_experiments` and spits out a clean, structured summary, leaving the messy UI navigation behind.
DVC MCP Server: Get Project, List Projects, and Audit Views
Before, verifying if a new repository was correctly linked required checking the project settings, then manually checking the user's permissions, and finally cross-referencing the dashboard view to ensure the correct metrics were visible. It was a multi-step, brittle process.
Now, your agent handles it. You ask it to verify the setup, and it runs `list_projects` and `get_user` to confirm scope. If you need to know the view's definition, it runs `get_view` and gives you the structured output. Everything is traceable and conversational.
Common Questions About DVC MCP
How do I use `list_projects` to see all my DVC workspaces? +
Running list_projects immediately returns all registered project names in your DVC Studio account. It's the quickest way to see the full scope of your ML work.
What is the difference between `list_experiments` and `get_project` using DVC? +
list_experiments shows the history—all the individual model runs and their metrics. get_project gives you the static metadata about the project itself, like its name and organization scope.
Can I use `get_view` to check a dashboard's settings? +
Yes. get_view retrieves the detailed, structural UI configuration for a specific dashboard. This is how you check if the view is set up correctly, even if the data is broken.
If I run `get_user`, what information do I get? +
The get_user tool pulls the profile details for a specified user. This helps you confirm the identity and scope of the person running the ML pipeline.
Do I need to run `list_views` before I can use `get_view`? +
It's helpful. First, use list_views to get the exact names of all available dashboard layouts. Then, feed one of those names into get_view to get the specific settings.
How do I check my permissions using the `get_user` tool? +
The get_user tool provides details on the token holder's authorized scope and role. It confirms exactly what data your AI client can access across the organization.
What happens if I try to list projects that don't exist using `list_projects`? +
The system returns a clear error message detailing the non-existent identifier. This helps you quickly correct the project name or scope in your prompt.
Which tool should I use to see all the specific metrics captured during an experiment epoch? +
You should use the list_experiments tool, which allows you to retrieve complex structural arrays. This shows precisely which metrics were logged for any given run.
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.
Use it with your favorite AI tools
Connect this server to Cursor, Claude, VS Code, and more.
More in this category
PlanetScale
Provision, branch, and manage serverless MySQL databases dynamically via AI.
BugHerd
Manage visual feedback and bug reports via BugHerd — track projects, tasks, and users directly from any AI agent.
Exponential Smoothing Engine
Apply Simple Exponential Smoothing (Holt-Winters family) for local deterministic time-series forecasting.
You might also like
AcoustID
Identify songs by audio fingerprint — like Shazam for developers. Search recordings by name, artist or MusicBrainz ID.
Langflow (Visual Multi-agent Orchestrator)
Orchestrate multi-agent AI workflows visually — execute flows, manage projects, and trigger webhooks directly from any AI agent.
IBAMA Dados Abertos
Access Brazilian environmental open data — query datasets, inspect resources, and analyze environmental metrics directly from IBAMA's official portal.