Explore Anthropic’s model platform as a component in an application, with task-specific testing and defined tool access.
Where it fits
The Claude platform documents model access and application features, including tools. Treat the currently documented capabilities as options to evaluate against the actual task and account.
A sensible first task
Use an approved sample document and ask for a short, source-grounded extraction or explanation. Keep the test cases and expected answers fixed when comparing configurations.
What to check
Review the actual account, feature, and output for the task you intend to do.
- Supported models, input formats, endpoints, and current feature limitations.
- Credential handling, data terms, usage limits, and costs.
- Behavior on missing information, invalid input, and unavailable tools.
How it connects to the bigger picture
Changing the provider does not remove the need for good context and review. A portable evaluation set helps distinguish a real improvement from a more persuasive response.
Sources and verification
Introduction to Claude · Anthropic