Nyckel

Nyckel is an AI platform for reliable predictions from examples. Classify images, text and structured data with stable API functions and usage-based plans.

AI Agent Development AI Developer Tools AI API Design

Visit Website

Overview

Nyckel is an AI platform for building reliable predictions from examples. It is designed for application workflows that need consistent decisions rather than open-ended text generation, such as spam detection, fraud detection, classification, and other repeated prediction tasks.

The product exposes AI decisions through stable functions and supports image, text, and structured data inputs. Its pricing page describes a usage-based model with a Free plan, a paid Starter plan for custom-trained models, a Business plan for automation and review workflows, and an Enterprise path for large-scale or regulated use.

Features

Automatic model selection

Tests hundreds of machine learning models against your data and automatically selects the best-performing model for each function.

Active learning for iteration

Surfaces suggestions that help improve accuracy over time and reduce manual annotation work.

Function-based API

Exposes one AI decision per stable API endpoint, making it suitable for repeated application calls.

Multiple data types and tasks

Supports image, text, and structured data workflows, with examples such as classification, detection, segmentation, and prediction.

Prediction output with confidence

Returns confidence scores with predictions so applications can use certainty in their own logic.

Hosted deployment

Hosts the models and handles server costs, scaling, and GPU optimization for teams that do not want to manage infrastructure themselves.

Use Cases

Image classification
Classify images into categories such as product photos, defects, species, or medical imagery using a model trained on your examples.
Text classification
Route incoming messages, detect spam or scams, or label sentiment and intent in support and review workflows.
Structured prediction
Work with tabular or event data for tasks such as fraud detection, churn prediction, anomaly detection, risk scoring, lead scoring, or demand forecasting.
Decision workflows with review
Use confidence scores, prediction review, and human review queues to support operational systems that need oversight before automation.
Prototyping to production
Start with the free tier for testing, then move to paid plans when you need custom-trained models, private inference, or higher-volume usage.

Pros and Cons

Pros

Built around repeated decision-making use cases where consistent outputs matter.
Supports multiple input types, including images, text, and structured data.
Includes confidence scores and monitoring/retraining-oriented workflows.
Offers hosted models, which reduces infrastructure work for teams.
Provides a clear usage-based pricing structure with a free entry point.

Cons

The public content does not include detailed integration documentation or request and response examples.
The source material does not explain setup steps, model limits, or advanced configuration in depth.

FAQ

What is a function in Nyckel?

Nyckel provides a stable API endpoint for a single AI decision, which it calls a function. The source material shows that functions can answer questions such as product identification in an image or whether a message is spam.

What is the difference between Free and Starter?

The pricing page says the Free plan gives access to AI models that are ready instantly, while Starter adds custom-trained models that improve automatically from real usage and feedback.

What does Nyckel return when a function is invoked?

The pricing page says invokes are prediction requests. When your application sends data to a function, Nyckel returns a prediction and a confidence score between 1 and 100.

Can I change plans later?

The source says you can upgrade or downgrade at any time from account settings, and upgrades take effect immediately.

Is there a way to get help before signing up?

The source shows a schedule-a-call path during signup, so users can contact the team if they need help getting started with a project.