REST API
A full-featured interface with 110+ secure endpoints to automate and integrate Keylabs into any corporate stack.
Infrastructure
Integrate the platform into your own AI infrastructure, with no data migration and no process changes.
Technology stack and integrations
Python SDK
Keylabs' own package to manage projects, files, assignments, and annotation export straight from your code.
Cloud storage
Native support for AWS S3, Google Cloud, Azure, and local storage through secure temporary links, so nothing is uploaded to the platform.
AI / ML models
Built-in zero-shot SAM-family models and smart tools for instant, automatic segmentation and tracking in 2D and 3D.
Roboflow
Connect your custom vision models through the Roboflow ecosystem to pre-automate markup of specific objects.
OCR (Google Vision)
Integrated Google text recognition to read and digitize handwriting, numbers, or documents while you annotate.
Multilingual localization
Set the interface language and use a built-in AI translator for text attributes, ideal for international teams.
Smart export
A high-performance export CLI for background delivery of datasets in COCO, YOLO (with tracking and attributes), or native Keylabs format.
Works with your data
AWS S3
Work with data directly through secure presigned URLs, with no need to copy gigabytes to third-party servers.
Google Cloud Storage
Fast sync with GCP so your team can start annotating the moment storage is connected.
Azure Blob Storage
Optimized enterprise access to Azure object storage with automatic rights control and request logging.
Local storage
Connect internal infrastructure to work securely with data that can't leave your perimeter.
Works with your models
SAM 3
One-click, high-accuracy segmentation of objects of any complexity.
YOLO26
High-speed bounding boxes and object tracking across thousands of frames at once.
Qwen2.5-VL
Multimodal tagging, video description, and question generation for visual data.
LayoutLMv3
Structure-aware markup of document text, tables, and forms for document AI.
SAM 3D
Automatic segmentation of objects in 3D space and point clouds.
Autodistill
Trains lightweight models from large architectures for task-specific pre-labeling.
RLHF
Human-feedback tools to fine-tune and align model responses.
BYOM (Bring Your Own Model)
Connect virtually any custom ML model through the platform's flexible API.
Customization for your project
Custom classes
Create unlimited markup shapes, from simple bboxes and polygons to skeletons and 3D cuboids.
Smart attributes
Configure dependent logic and any field type for detailed object descriptions.
Flexible tags
Project tag dictionaries for fast classification, mass labeling, and instant search across large datasets.
Configurable workflows
Build multi-stage task chains with their own colors and assigned performers.
Automatic validation
Run custom QC scripts with a UI debugger that blocks file transfer when errors appear.
For which scenarios
Enterprise AI teams
Connects to existing corporate infrastructure with full compliance to strict security standards.
Data operations teams
End-to-end pipeline automation with fast performer management and control of budgets and deadlines.
ML engineers
Launch pre-labeling with advanced VLM and foundation models, then export datasets via a convenient CLI.
AI vendors
Full white-label customization under your own brand, scaling to dozens of external customers.
