AI Agent Orchestration

DLinkClaw has a built-in AI agent orchestration engine. Describe a flow in natural language and the agent will orchestrate it for you.

Quick Creation

In the Agents page, click “New” and type, for example:

Every morning at 9 AM, summarize yesterday’s sales data, draft a weekly report, post it to #reports, and notify the sales director.

DLinkClaw parses it into: schedule trigger → pull data → AI generation → publish to channel → notify stakeholders.

Agent Capabilities

• Intent understanding: extracts triggers, actions, and parameters from natural language.
• Tool calling: integrates 40+ built-in tools (databases, IM, email, tickets, APIs).
• Multi-step planning: breaks complex tasks into executable plans.
• Human approval: inserts approval steps at critical nodes for control.

Access Control

Each agent can be bound to independent tool permissions and data scopes, following least-privilege principles.

Example: Ticket Auto-Triage

agent: ticket-triage
tools: [ticket_system, knowledge_base, notify]

prompt: >
  Based on the ticket content and knowledge base matches, assign a priority (P0-P3),
  assign it to the right owner, and announce it in #triage.