Nyckel
Managed prediction APIs for classification, detection, segmentation, and search, with models that learn from reviewed examples.
About Nyckel
Nyckel is a managed machine learning platform for teams that need prediction APIs without operating their own training and serving stack. A team can define a classification, detection, segmentation, or search function, add examples, and call the result through a production REST API. The service evaluates candidate model families, selects and trains a suitable model, hosts it, monitors it, and can improve it as reviewed predictions and corrected samples arrive. The console supports an initial no-code workflow, while developers can integrate functions through the API and retain an audit trail with captured invocations. Pricing is usage based: the Free plan includes 100 monthly invokes, 200 samples, and 5 functions; paid tiers add custom training, private inference, review tooling, automation, higher limits, and lower unit rates. This approach reduces infrastructure work, but teams still need labeled examples, a review process, and a budget for usage that exceeds each plan’s allowance.
Best for
Best use cases
Prototype a production API on the Free plan before committing to a paid tier.
Train custom models from production examples and corrections on Starter.
Run high-volume automated decisions with review and export controls on Business.
Compare before choosing
Free has tight limits but provides a production API and no monthly fee.
Starter adds custom training and private inference at $149 per month.
Business costs $599 per month and trades a higher base price for much larger allowances and lower unit rates.
Features
Custom classification
Train text, image, or structured-data classifiers from labeled examples, then serve predictions through a managed REST API.
Visual detection and segmentation
Build functions that locate or segment objects in images without selecting, deploying, or maintaining the underlying model architecture.
Feedback-driven operations
Capture production invocations, review predictions, add corrections, and let the managed ML engine retrain as the dataset changes.
Pros
- The $0 Free plan includes a production API, so teams can test integration before paying a monthly fee.
- Nyckel manages model selection, training, hosting, and monitoring, reducing the ML infrastructure a product team must operate.
- Captured invocations, prediction review, and feedback-based improvement connect production use with an ongoing correction workflow.
Cons
- Free usage is limited to 100 invokes, 200 samples, and 5 functions per month, which is restrictive for sustained production traffic.
- Overage charges apply after included usage, so cost grows with invocation and sample volume.
- The managed model-selection workflow gives teams less direct control over architecture choice and training configuration.
Pricing Plans
Free
- AI models ready instantly
- Production API
- Classification, detection, and segmentation
- Prediction review
- Invoke capture for real-world testing
- 100 invokes per month included
- Up to 200 samples
- Up to 5 functions
- $0.005 per invoke after included usage
Starter
- Everything in Free
- Custom trained models
- Improve accuracy with feedback
- Smart review suggestions
- Private inference
- 15,000 invokes per month included
- Up to 30,000 samples
- Up to 20 functions
- $0.005 per invoke
- $0.0025 per additional sample
Business
- Everything in Starter
- Decision thresholds and automation
- Human review queues
- Multi-model decision pipelines
- Prediction auditing and export
- Larger prediction retention
- 300,000 invokes per month included
- Up to 600,000 samples
- Up to 100 functions
- $0.001 per invoke
- $0.0005 per additional sample
Enterprise
- Everything in Business
- Dedicated infrastructure
* Pricing information is for reference only. Please refer to the latest announcement on Nyckel's official website.