Skip to main content

Autonomous Agents

Autonomous Agents are the "brains" of Arrotech Hub. Unlike static scripts, an Agent is a stateful entity that uses LLMs to reason about tasks, select tools, and execute workflows autonomously.

Architecture Overview​

Arrotech agents follow a Perceive-Plan-Act loop:

graph TD
    A[Trigger: Schedule/Event] --> B[Perceive: Input Analysis]
    B --> C[Plan: Tool Selection]
    C --> D[Act: MCP Execution]
    D --> E{Success?}
    E -- No --> F[Retry Logic]
    E -- Yes --> G[Update State]
    F --> C

Key Capabilities​

  1. Stateful Memory: Agents can recall results from previous steps within a single execution sweep.
  2. Dynamic Tooling: Powered by MCP, agents automatically discover "Skills" based on your Connections.
  3. Universal Guardrails: All agent actions are filtered through a safety layer to prevent unauthorized data access.

Advanced Usage​

Scheduling​

You can schedule an agent using standard Cron expressions or predefined intervals. This allows for background tasks like "Generate a daily sales report every morning at 8 AM".

Manual Execution​

If you need to trigger an AI agent as part of an external application, use our high-performance manual execution endpoint.

[!important] Agents require at least one active Connection to perform actions. Without a connection (e.g., Slack or HubSpot), the agent will only be able to perform pure reasoning tasks.

Learn More​