cvPal
cvPal
Dashboard

Manage credentials
and connection settings

Core Concepts

MCP Server

cvPal implements the Model Context Protocol (MCP) — an open standard that gives AI assistants structured access to external tools and data. Instead of writing scripts or using platform-specific APIs, your assistant calls cvPal tools through natural language.

How it works

MCP defines a universal interface between AI models (Claude, GPT-4, Gemini, etc.) and external capabilities. When your assistant connects to the cvPal MCP server, it discovers all available operations and can invoke them on your behalf. The protocol handles authentication, parameter validation, and result formatting automatically.

cvPal exposes three types of capabilities through MCP:

Tools

Executable functions your AI assistant can call — merge datasets, export to YOLO format, augment training data, and more.

Resources

Read-only data like dataset reports, class distributions, and project metadata available on demand.

Prompts

Pre-built templates that guide your assistant through complex multi-step workflows like dataset analysis.

Connecting to the Server

cvPal uses the Model Context Protocol with Streamable HTTP transport. Each user receives a unique, authenticated connection endpoint — never shared publicly.

How to retrieve your connection details:

  1. Sign in to your cvPal dashboard
  2. Navigate to Settings → Credentials
  3. Copy your personal endpoint URL and bearer token

Ready to connect?

Follow our platform-specific guides for Cursor, Claude Desktop, VS Code, and other MCP clients.

Go to Integrations