Skip to main content
Make a direct Adapt-1 request and use the JSON result in your application. Choose cURL, Python, or JavaScript below; all three send the same request.

Set your API key

Use an Adapt-1 Unit API key, stored in your local environment or server-side secret store:
Environment
The examples use https://rei-neuroadapt-api.reilabs.org/api/v1. Keep the key out of browser code, source control, and logs.

Make your first call

This request asks the State gateway for available evidence. It is read-only: update_memory_state is explicitly false.
Python uses the standard library. Run JavaScript as an .mjs file with Node.js 20 or later. Both examples run on the server.

Read the result

An empty evidence result is valid when nothing relevant has been stored. This is an illustrative minimal response, not a captured live result; the service can return additional fields.
Empty-evidence illustration
Your code consumes these fields directly. Generated language is not required.
The hosted service resolves the effective identity from the bearer token. The body field satisfies this request’s compatibility contract; changing it does not select another tenant or create isolated state. Use a separate credential for a separate identity, or a separate domain_id for a task-scoped Domain.

Add a learning loop

The first call checks authentication and the response boundary. Next, run a complete learning example: create a fresh Domain, submit observations, inspect what was admitted, and query using retained state.

Build your first learned result

One complete Python file, included observations, before/after responses, and a read-only reuse step.

Choose a workflow

Start from the prediction, decision, evidence, or control result your application needs.

Build with an AI agent

Give your coding agent the request contract and an implementation checklist.

If the call fails

For exact request boundaries and schema coverage, see HTTP and request contracts.