The same agent keeps its identity and shared memory across connected applications. Chat, browser or your own app: switching applications does not create a separate brain. This page explains the technical integration for your own agents, not how to order Employee.
An Agent-ID identifies the agent. Its connected applications use the same identity and shared memory. Each application only receives the access rights it needs; a shared brain does not mean unrestricted access.
For example, a decision saved in chat can be available to the same agent in an internal app. Both connections must point to the same Agent-ID and memory pool, with the relevant permissions. We agree which connections your project needs.
Which agent does the application connect to, and who is responsible for it?
Which information may the application read or store, and which actions may it trigger?
Which decisions and actions need a record that can be checked later?
Recall retrieves information relevant to the current task from the agent's brain.
Validity checks address whether information is current and may be used in this application.
New decisions and corrections go back into the same shared memory.
Which application do you want to connect? Tell us about the agent, the information it needs and the actions it should be allowed to take. Access and terms are agreed individually.