Agent¶
This section groups API operations for managing agents, binding resources, invoking runs, automating work, and configuring the runtime. Open an action page for the required identifiers and call parameters.
Choose a task¶
Task |
Page |
Description |
|---|---|---|
Manage agents and versions |
The agent resource interface is used to manage agents, their versions and resources, models, operating strategies and operating environment bindings. |
|
Choose an invocation protocol |
Before calling an agent, obtain its Agent Card; then you can use the A2A or MCP protocol to initiate the call. |
|
Manage sessions and read messages |
The session interface is used to manage agent session shells and their presentation metadata. |
|
Review run feedback |
The feedback interface is used to review saved agent running feedback and summary statistics; submitting feedback is still completed interactively during the agent runtime. |
|
Configure and run automation tasks |
Automation tasks associate configured agents with triggering methods. |
|
Prepare and confirm candidates |
The agent building interface is used to read creative resources, generate candidate solutions, and confirm or cancel candidate solutions. |
|
Configure channel connections |
A channel instance holds a connection configuration and credential reference for a specific Provider. |
|
Configure agent models |
The agent model configuration saves non-sensitive metadata such as model selections, capabilities, and connection references. |
|
Configure runtime policies |
The run policy configuration is the control plane metadata used by the agent when it is run. |
|
Manage knowledge-base metadata |
List, retrieve, create, and update the knowledge bases attached to agents. |
|
Manage tools and tool connections |
The tool interface is used to manage workspace tools, query reference relationships, and manage built-in GitHub, Grafana, and email tool connections. |
|
Manage skills and versions |
The skills interface is used to manage workspace skills, import packages, select versions, and submit run requests. |
|
Configure callback rules |
List, retrieve, create, update, and delete callback rules. |
|
Define node invocation templates |
The agent task template describes the messages, inputs, and context configuration used by the workflow node when invoking the agent. |
|
Bind workflow nodes to templates |
Agent workflow binding associates workflow nodes to an agent task template. |
|
Choose a runtime provider |
List runtime providers and retrieve a provider profile. |
|
Configure system agents |
Retrieve or update system agent configuration, including the available GitHub projects used by the configuration. |
|
Load or export agent packages |
Load an agent package into the workspace or export an existing package. |
|
Inspect runtime records |
Inspect data partitions, runtime tasks, task events, manifests, and round snapshots for an agent run. |
|
Track asynchronous resource operations |
List and inspect agent resource operations, or cancel an operation that can be cancelled. |
|
Connect external services |
List, retrieve, create, and update agent connections, and probe MCP connections or create MCP tools in batches. |