About Nyckel

Nyckel is a no-code AutoML platform founded in 2021 by George Mathew and Oscar Beijbom (Y Combinator W22 batch), headquartered in Oakland, California. It enables developers and businesses to build, train, and deploy custom machine learning classification functions without requiring ML expertise or infrastructure management. The platform supports three input data types: images, text, and tabular data. For each, Nyckel automatically evaluates hundreds of ML models against the user's labeled samples, selects the best-performing one, and deploys it to an elastic inference API with claimed 99.99% uptime and under 300ms latency. The entire cycle from labeled data to a live production endpoint can take minutes. Nyckel's function types include image classification, text classification, image/text tagging, semantic image search, object detection, and tabular classification. It also offers a library of pretrained public classifiers that users can invoke immediately without any training data. Active learning features surface annotation suggestions to improve model accuracy over time with minimal manual labeling effort. The platform targets small-to-mid-sized companies in eCommerce, social media, content moderation, and SaaS who need reliable predictions without a dedicated ML team. It is SOC2 and HIPAA compliant. As of 2026, Nyckel remains an active, independent private company with approximately 3 employees, and shows no signs of acquisition or shutdown.

AI Agent Use Cases

- AutoML: automatically benchmarks hundreds of ML models on user data and deploys the best one

- Multi-modal inputs: image, text, and tabular classification in a single platform

- Object detection: annotate images in-browser and deploy a detector in minutes

- Semantic image search: search an image library using images or natural-language text queries

- Pretrained public classifiers: ready-to-use classifiers requiring zero training data

- Active learning: platform surfaces labeling suggestions to improve model accuracy continuously

- REST API deployment: trained models exposed as versioned REST endpoints with elastic scaling

- SOC2 and HIPAA compliance for handling sensitive or regulated data

Available Actions

These are the specific actions that AI agents can perform with this tool

Classify Image

5 inputs

Classifies an image by applying Nyckel's pre-trained machine learning models to identify its content or category.

Inputs

functionId

Select a function or provide a custom function ID.

labelCount

The number of labels to return, along with their confidences. When not specified, only the highest confidence result is returned.

includeMetadata

Whether to include the label metadata in the response.

capture

Whether to enable invoke capture for this invoke. Invoke capture saves informative samples captured from your invokes for future annotation.

externalId

Your unique identifier for this sample. This will be used if this invoke is saved through Invoke Capture.

Classify Text

6 inputs

Classifies input text using Nyckel's pre-trained models to identify categories or labels relevant to the content.

Inputs

functionId

Select a function or provide a custom function ID.

textInput

The text input to classify

labelCount

The number of labels to return, along with their confidences. When not specified, only the highest confidence result is returned.

includeMetadata

Whether to include the label metadata in the response.

capture

Whether to enable invoke capture for this invoke. Invoke capture saves informative samples captured from your invokes for future annotation.

externalId

Your unique identifier for this sample. This will be used if this invoke is saved through Invoke Capture.

Extract Text from Image

2 inputs

Extracts and returns the textual content embedded within an image using its URL.

Inputs

functionId

Select a function or provide a custom function ID.

includeRegions

When set to true, return the regions of the image that contained text