> ## Documentation Index
> Fetch the complete documentation index at: https://docs.loqate.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Usage

> Discover and add the Reach agent in watsonx Orchestrate, then use it in chat, as a collaborator, or as a step in an agentic workflow.

<Note>
  **Coming soon.** The Reach agent for IBM watsonx Orchestrate is not yet generally available. This documentation describes the agent ahead of its launch — features and steps may change before release.
</Note>

The Reach agent is discovered and added from inside watsonx Orchestrate — there's nothing to install locally. Once added, you can use it three ways, all returning the same structured [verification report](/ai-agents/wxo/overview#structured-output).

## Add the agent to your instance

Reach is published as a native agent in the watsonx Orchestrate agent catalog. From your instance, browse the catalog, select **GBG Reach**, and add it. The **Loqate API key** the agent uses to authenticate is issued when you provision and purchase Loqate through watsonx Orchestrate — there's no separate Loqate account to set up.

<Note>
  Discovering and adding a native agent is a standard watsonx Orchestrate workflow. For the exact steps in your environment — including the agent catalog and connection credentials — follow the [watsonx Orchestrate documentation](https://www.ibm.com/docs/en/watsonx/watson-orchestrate/base), using the version that matches your instance.
</Note>

## 1. Interactive chat

The simplest pattern — talk to the agent directly. It detects the data type, runs the right verification, and returns a decision with a confidence score.

```text theme={null}
You:  Verify this email: jane.doe@example.com

Reach:  jane.doe@example.com → accept (confidence 0.95, policy: standard)
        Deliverable, format valid, no risk flags.
```

You can also paste a block of records, or **upload a CSV**, and ask the agent to verify every row in one go.

## 2. Agent-to-agent (collaborator)

Add Reach as a **collaborator** of another agent so a supervisor agent can hand off verification mid-conversation. The supervisor stays in control of the dialogue; when the user asks to verify, validate, or cleanse an address, email, or phone number, it delegates to Reach and receives a structured result back.

This is configured in watsonx Orchestrate by listing Reach among the agent's collaborators — no code required.

## 3. As part of an agentic workflow

For bulk, scheduled, or headless work, use Reach as a step inside a watsonx Orchestrate flow. A typical contact-cleansing flow reads records from your system of record, verifies and standardises each one through Reach, then writes the cleansed records back:

```text theme={null}
[Read contacts] → for each record → [Reach: verify & standardise] → [Write cleansed records]
```

Reach supplies the verification step in the middle; the read and write-back steps connect to your own systems (CRM, data store, spreadsheet). Reach's structured output — decision, confidence, standardised fields, and reason per record — flows straight into the next step.

## Policies

Every verification is evaluated against a policy that sets the accept / review / reject thresholds. Reach ships with `strict`, `shipping`, `standard` (the default), and `permissive`. You can ask the agent which policies are available or to explain a specific one, and you can name a policy in your request (for example, "verify these addresses using the strict policy").

<Tip>
  See [Policies](/ai-agents/reference/policies) for the full breakdown of each policy's thresholds and when to use it.
</Tip>