cvPal Documentation

cvPal is a managed MCP server that turns your AI assistant into a computer vision engineer. Manage datasets, transform labels, merge and augment training data, import from external platforms, and export to 20+ formats — all through natural language.

Why cvPal?

Traditional CV dataset pipelines require custom scripts, manual data wrangling, and platform-specific tooling. cvPal replaces that friction with a single MCP endpoint your AI assistant connects to — giving it 15 purpose-built tools that handle the entire dataset lifecycle.

cycle

Full Dataset Lifecycle

Create, clean, augment, merge, label, and export datasets across 10+ annotation types including detection, segmentation, and keypoints.

cloud_download

Multi-Platform Import

Search and ingest datasets directly from Roboflow, Kaggle, and Hugging Face — no manual downloads or format conversions.

memory

Serverless GPU Processing

Heavy compute jobs run on scalable infrastructure. Track progress in real-time and retry failures automatically.

Ready to start?

Install the MCP server, connect your AI assistant, and run your first tool call — the entire setup takes minutes.

Quickstart

Connect Your AI Assistant

cvPal works with any MCP-compatible client. Sign in to your dashboard to retrieve your personal connection details, then follow the integration guide for your preferred platform.

Get Connection Details View Integrations

cvPal in action — searching, inspecting, and importing datasets through natural language

Explore the Docs

Getting Started

Connect your AI assistant to cvPal and run your first computer vision task in minutes.

Integrations

One-click setup for Cursor, Claude Desktop, VS Code, Gemini CLI, OpenAI SDK, and more.

MCP Server

Understand how the Model Context Protocol connects your AI assistant to cvPal's tools.

Dataset Management

Inventory, organize, and manage datasets with full version history and branching.

Label Operations

Rename, remove, and audit label classes across classification, detection, segmentation, and more.

Processing

Merge, clean, augment, and export datasets to 20+ formats on serverless GPU infrastructure.

External Sources

Discover and import datasets from Roboflow, Kaggle, and Hugging Face into your workspace.

Jobs & Reporting

Monitor async processing pipelines, generate dataset analytics, and recover from failures.

Complete Tool Reference

Every tool available through the cvPal MCP server. Instant tools return results immediately; async tools run on serverless infrastructure with progress tracking.

list_user_datasetsInventory all datasets with metadataInstant

delete_datasetRemove a dataset and all versionsInstant

modify_labelsRename or remove label classesAsync

query_datasetCount labels or locate imagesInstant

merge_datasetsCombine datasets with class remappingAsync

clean_datasetStrip unlabeled items from a datasetAsync

augment_datasetApply transforms to expand training dataAsync

export_datasetExport to YOLO, COCO, VOC, and 20+ formatsAsync

search_external_datasetsFind datasets on Roboflow, Kaggle, HFInstant

inspect_external_datasetPreview metadata before importingInstant

ingest_datasetImport an external dataset into cvPalAsync

report_datasetGenerate distribution and quality analyticsInstant

get_processing_statusPoll progress of any async jobInstant

retry_jobRetry or resume a failed pipelineAsync

get_helpRetrieve docs and code examples via MCPInstant

Questions or feedback?

Our team is here to help with integration issues, tool behavior, or feature requests.

Contact support →