This section explains the building blocks of the Xians Agent SDK — what each feature does, why it exists, and how to use it. Each page is written to be readable on its own, so you can jump straight to the feature you need.
A Xians Agent is a deployable unit that owns everything it needs to operate: workflows that run its logic, knowledge that guides its behavior, documents that store its data, and messaging channels that connect it to users and other agents.
graph TB
subgraph Agent
W[Workflows<br/>run the logic]
K[Knowledge<br/>prompts & instructions]
D[Document DB<br/>structured data]
S[Secret Vault<br/>credentials]
end
U[Users] <-->|Messaging| W
E[External Systems] <-->|Webhooks| W
O[Other Agents] <-->|Cross-Agent Workflows| W
T[Time] -->|Schedules| W
H[Humans] <-->|HITL Tasks| W
Here's how the features combine in a real scenario — an autonomous content agent:
sequenceDiagram
participant S as Schedule
participant W as Workflow
participant K as Knowledge
participant D as Document DB
participant O as Other Agent
participant H as Human
participant M as Messaging
S->>W: Trigger daily content check
W->>K: Fetch instructions
W->>D: Get pending content items
W->>O: Delegate analysis (Cross-Agent Workflow)
O-->>W: Return analysis
W->>H: Create approval task (HITL)
H-->>W: Approve with edits
W->>M: Notify user of publication
W->>D: Update content status
Schedule triggers the workflow every morning — no human needed to start it.
Knowledge provides instructions that can be updated without redeploying code.
Document DB holds the content items being processed.
Cross-Agent Workflows delegates analysis to a specialized agent.
HITL pauses for human approval before anything is published.