References

Changelog

Version history and release notes for cvPal's MCP server.

v1.0.0Current

MCP Server Release

February 2026

Tools

Unified external sources with search_external_datasets, inspect_external_dataset, and ingest_dataset
Added modify_labels tool for removing/renaming label classes
Added query_dataset tool for instant label counting and image searching
Added report_dataset, get_processing_status, and retry_job tools
Added get_help tool for documentation and examples
Processing tools: merge_datasets, clean_dataset, augment_dataset, export_dataset
Dataset management: list_user_datasets, delete_dataset

Integrations

Support for Cursor, Claude Desktop, Claude Code, Gemini CLI
OpenAI SDK and Codex CLI integration
VS Code MCP support
Streamable HTTP MCP endpoint

Infrastructure

Serverless GPU processing on RunPod
Content-addressable storage for datasets
Async job system with status polling
Automatic dataset type detection (classification, detection, segmentation, keypoints, OCR, video, point clouds, vision-language, image-to-image)

Export Formats

Detection: yolo, coco, voc, kitti
Segmentation: yolo_seg, coco_seg, masks
Keypoints: coco_keypoints, yolo_pose
Classification: folder, csv
Vision-Language: json, caption, vqa
Image-to-Image: paired, pix2pix
Video: tracking
3D: ply
Hub: huggingface

Connecting to cvPal

Connection details (endpoint URL and authentication token) are provided per-user through the dashboard. See the Integrations page for setup instructions.