Skip to main content
Version: 11.5.0

Self-Managed Cloud Agents

A self-managed agent is an agent harness you run in your own infrastructure, such as a container or your own cloud account. HYPR AgentPass™ governs it once its traffic goes through the gateway.

A harness in your own container or cloud account sends model calls to an AI provider connector and tool calls, with a workload credential, to an MCP or REST API connector. The gateway forwards them to the LLM provider and to your tools and APIs. On the Agents page the harness runs in the cloud and acts as a workload when no person can authorize it.

Route Its Calls Through the Gateway​

A harness makes two kinds of call. Each goes through its own kind of connector.

  • Model calls. Point the harness at an AI provider connector instead of the provider. AI provider connectors carry large language model (LLM) traffic from desktop and cloud agents.

  • Tool calls. Point it at a Model Context Protocol (MCP) server or REST API connector. Those connectors expose your tools to agents.

See Connectors for each kind.

Give It an Identity​

Whose access the harness uses decides the identity it carries.

  • No person is available to authorize it. Give it a workload agent identity. It acts with a role of its own, only on the connectors you name. On the Agents page it is listed under Acts as: Workload.

  • A person runs it and can authorize it. Use the ordinary path, where that person authorizes the connector once. See Connecting to a Gateway MCP Connector.

A workload agent's Runs setting decides where the Agents page lists it. Choose In the cloud for a harness in your cloud account. See Creating a Workload Agent.