Endpoint Response Model Generator: from an API response to a model
Connecting an external system usually starts with a JSON response and a blank page. Which fields become columns? What do you do with nested lists? Endpoint Response Model Generator reads the response from an endpoint test and proposes both parts of the answer: a data model for the records, and the field mappings for a sync workflow. It appears where you set up and test API endpoints in the integration settings.
Worknano is customisable with AI and focused on core functionality. Synced data lands in a normal model, with the same UI, UX and database as everything else, not in a special data store.
What it does
- Reads the response body of a tested endpoint, plus its name, path, method and slug when you provide them.
- Proposes a flat model. Useful single values become real columns, especially the ones you will report on or filter by.
- Recommends how to sync. A response with one object suggests updating one existing record. A list, or a common wrapper around a list, suggests creating records.
- Stores nested lists or objects as JSON text only when there is no better option.
- Starts the model with an id and ends it with a synced time, and maps only the fields it generated.
- Handles custom fields that the provider sends as name, type and value entries, when the target model has a place for them. The import then adds the definitions to the parent record and the values to each row.
When it helps
- A CRM import. You test the endpoint that lists deals in another tool. The agent proposes a deal model with amount, stage and close date as real columns.
- A single settings object. An endpoint returns one object with your plan details. The agent suggests a model and an update rule, not a new row each time.
- A project tool with custom fields. Each project in the source defines its own fields. The agent maps those so the values arrive with their definitions.
How it works with your data
This agent has no tools. It does not call the endpoint itself and it does not write records. The endpoint test already ran, and the response is passed in. The agent returns a model draft and a mapping suggestion, and you review both before anything is saved. Applying a model is a configuration change, so it stays with the people who may edit configuration. Credentials and connection secrets are never part of what the agent sees.
It uses a fast model. The input is a response body and the output is structured, so speed matters more than depth.
What it does not do
- It does not test or call the endpoint. You do that first, and a response with no useful fields gives a thin proposal.
- It does not guess meaning that the response does not show. Check names and types, especially for dates and money.
- It does not run the sync. The workflow does, once you have accepted the mapping.
For a hand-written model from a description, use Model Designer. For the sync step itself, see Workflow Creator.
Make it yours
This agent is a record in the Agents module. You can edit its instructions to match your naming rules, change the model or temperature, duplicate it for one particular provider, or start from a blank agent.
Start with Build your own agent, read How to edit a page assistant and a reusable agent for the editor steps, or see the bigger picture in AI agents and AI assistants.