Destination actions

Besides having your agents run queries against your Harbor, you may also require your agents to either look up data from or write data to live external systems in real-time.

For example, you may want to enable your users to instruct their agents to update or look up certain fields in Salesforce or other systems on the fly.

Destination actions are policies you can specify that dictate exactly what fields in external systems the agent is allowed to read and/or write.

You can specify action policies for your connecting AI agents by going to the Destination actions tab in your Harbor.

In the example below, you can see two types of action policies specified. With these two policies there is a guarantee that, in addition to being able to query data on this Harbor, all agents going through this Harbor would also be allowed to:

• Update a specific list of fields on the Salesforce Account object.
• Update the Next Step field on the HubSpot Deals object.
• Look up HubSpot Deals but only get back these fields:
  • Amount
  • Close Date
  • Deal ID
  • Deal Name

Any attempt to interact with data in external systems outside of this scope would be rejected.

Security and permission guarantee

Harbor destination actions provide a security and permission guarantee for AI agents interacting with live systems. Through them, you can guarantee with absolutely certainty that an AI agent would not be able to update or read from a field in any system that is outside of the allowed scope you specify.