Using Sendoso in Databricks Genie One

Introduction

The Sendoso MCP is available in Databricks Marketplace. Once your workspace admin adds it, you can ask Genie One to find gifts in the Sendoso marketplace, research recipients, and send gifts, all without leaving Databricks. Your admin controls which Sendoso actions need approval through Unity AI Gateway.

Available on Core, Advanced, and Enterprise plans.

Listing: Sendoso MCP in Databricks Marketplace

Prerequisites

• Sendoso account: You need an active Sendoso login with MCP access. Ask your Sendoso account team to enable access for your organization before setup.
• Databricks workspace: Unity Catalog and Genie One must be enabled in your workspace.
• Workspace admin: An admin installs the listing and gives users access. Each user then turns Sendoso on and signs in for themselves.

Step 1: Add the Sendoso MCP from Databricks 

Marketplace (admin)

1. Open the Sendoso MCP listing and select Get access.
2. Install the connection into a catalog and schema in your workspace. [CONFIRM: the exact install dialog labels.]
3. Give users access to the connection: USE CATALOG on the catalog, USE SCHEMA on the schema, and EXECUTE on the Sendoso MCP service.

Select Get access on the Sendoso MCP listing

Step 2: Turn on Sendoso in Genie One

1. In Genie One, select Customizations, then Connectors.
2. Find your Sendoso connection and switch it on. It is listed under the name your admin gave it, in the format catalog.schema.name.

Each user turns the connection on for themselves. It stays off until you switch it on.

Select Customizations, then Connectors, then switch on your Sendoso connection

Step 3: Sign in to Sendoso

1. Select your Sendoso connection card, then select View in Catalog.
2. Select Login and sign in with your Sendoso account.
3. When sign-in is complete, the list of Sendoso tools loads on the page.

Your sign-in authorizes the Sendoso MCP for you only. Other users in the workspace sign in with their own Sendoso accounts.

Tip: If Genie asks you to sign in to Sendoso during a chat, go to the connection's page in Catalog and select Login.

Select View in Catalog on your Sendoso connection

Select Login and sign in with your Sendoso account

Step 4: Ask Genie

Mention Sendoso in your request so Genie knows which tools to use. For example:

• Using Sendoso, find three gift ideas under $50 for a customer marketing manager who just renewed.
• Using Sendoso, look up this contact in Salesforce and suggest a gift.
• Using Sendoso, send a $25 eGift card to [name] at [email] with a thank-you note.

Genie shows each Sendoso step it takes under Thought process.

Mention Sendoso in your request, then open Thought process to see each Sendoso step

Step 5: Review and approve a send

Every send happens in two steps: a preview, then a confirmation. Nothing is charged on the preview.

1. Genie shows 1 tool needs your approval. Review the details, then select Allow once to continue or Deny to stop.
2. Genie shows a send preview with the gift, total cost, recipient, note, and delivery method. Check each one, then reply to confirm.
3. Approve the send when Genie asks again.
4. Genie confirms the send and shows its Send ID.

Select Allow once to see the send preview

Check the gift, cost, recipient, and note

Select Allow once again to send

Genie confirms the send and shows its Send ID

Why Genie asks first: Sendoso actions that send, change, or cancel something are listed under Write/delete tools on your Sendoso connection's page in Genie One (Customizations, then Connectors, then select the connection). There they are set to Ask first, and each tool has its own setting next to it. Your admin can also require approval with a Unity AI Gateway policy.

Write/delete tools, including execute_send, set to Ask first

For admins: control what needs approval

Approval rules are policies on the Sendoso MCP service in Unity AI Gateway. Open AI Gateway, select the Sendoso MCP service, then open Policies to require approval for a tool, allow it, or deny it. 

Policies can differ by user or service account, so an approved automated flow can run without a person in the loop while everyone else still approves each send. [CONFIRM: menu path and per-principal policies, from Databricks documentation.]

Troubleshooting