Language:JapaneseEnglish

fiftyone Claude Code Plugin

Topics
AI Agents & AI App Development · Data Engineering & Analytics · Code Quality, Review & Testing
First cataloged
2026-07-09 (UTC)
Explanation last updated
2026-07-16 (UTC)
Source (GitHub) last updated
2026-07-21 (UTC) (37 days ago)

The explanation below is AI-generated. Please verify it against the sources.

This plugin is a bundle of AI-assistant skills for working with FiftyOne, an open-source platform for building and inspecting computer vision datasets and models. According to the README, the skills pair with a separate FiftyOne MCP server so that an assistant can carry out tasks such as importing or exporting datasets, finding near-duplicate images, running zoo-model inference, computing evaluation metrics, and visualizing embeddings from natural-language requests. Other skills in the package help with writing custom FiftyOne plugins, following FiftyOne's Python coding conventions, generating notebooks, triaging GitHub issues, driving the FiftyOne App through Playwright, and producing Data Lens connectors for external databases. Several of these skills, such as code-style guidance and plugin scaffolding, work without the MCP server. The homepage documentation describes FiftyOne itself as a broader product for data annotation, curation, and model evaluation with many third-party integrations.

About the service

FiftyOne is an open-source toolkit (with an accompanying Enterprise edition) for visualizing, curating, annotating, and evaluating computer vision datasets and models. Per the README, this skill collection reaches FiftyOne through a separately installed FiftyOne MCP server, which lets an AI assistant manipulate datasets and invoke FiftyOne's operators.

What you can do with fiftyone

  • Import or export datasets in formats such as COCO, YOLO, VOC, CVAT, and Hugging Face Hub through natural-language requests
  • Detect and remove near-duplicate images using brain similarity search
  • Run zoo models for detection, classification, segmentation, or embedding generation
  • Compute model evaluation metrics such as mAP, precision/recall, and confusion matrices
  • Visualize dataset embeddings in 2D to spot clusters and outliers
  • Scaffold custom FiftyOne plugins (operators and panels) and related UI components
  • Get assistance writing FiftyOne-conventional Python code, building notebooks, triaging issues, and troubleshooting common errors
  • Generate Data Lens connectors for external database schemas like PostgreSQL, BigQuery, or MySQL

Sources

Back to list