Demo: Multi-Agent Orchestration in Cursor

This project demonstrates how specialized AI agents can collaborate to solve complex, multi-step tasks within the Cursor IDE.

Project Goal

The primary goal is to showcase the orchestration capability of Cursor's agentic framework. By defining a high-level "Trip Planner" agent that coordinates with specialized "Flight" and "Hotel" researchers, we illustrate a modular and scalable approach to automated problem-solving.

Key Features

• Multi-Agent Collaboration: Transparent delegation of sub-tasks from a primary agent to specialized sub-agents.
• Skill-Based Execution: Integration of custom "Agent Skills" that execute shell scripts to perform specific actions (e.g., searching data).

Subagents and Agent Skills

Agents

• @trip-planner: Orchestrates the entire trip planning process by delegating tasks to specialized agents.
• @flight-researcher: Researches flight options based on origin and destination.
• @hotel-researcher: Researches hotel accommodations at the destination.

Agent Skills

• @research-flight: Executes flight search scripts to retrieve travel options.
• @research-hotel: Executes hotel search scripts to retrieve accommodation details.

Relationships across subagents and Agent Skills

graph TD
    TripPlanner[Trip Planner Agent] --> FlightResearcher[Flight Researcher Agent]
    TripPlanner --> HotelResearcher[Hotel Researcher Agent]
    FlightResearcher --> ResearchFlight[Research Flight Skill]
    HotelResearcher --> ResearchHotel[Research Hotel Skill]

Loading

Usage Instructions

To run the demonstration:

1. Open the Cursor IDE in this repository.

2. Open the Chat panel (Cmd+L or Ctrl+L).

3. Use the @trip-planner agent by typing its name or using the /trip-planner slash command (if configured).

4. Provide a prompt such as:

   I want to go to San Francisco from NYC. Please write a trip report in @reports/plan_to_san_francisco/trip_plan.md and logs in @reports/plan_to_san_francisco/task_logs.md.
   

5. Observe how Cursor orchestrates the sub-agents and generates the requested files.

Directory Structure

• .cursor/agents/: Contains agent definitions (YAML-frontmatter Markdown) defining their roles and capabilities.
• .cursor/skills/: Defines custom skills that agents can use, including the scripts they execute.
• reports/: The output directory for generated trip plans and task logs.
• docs/: Project documentation and assets, including screenshots.

Technical Implementation

The orchestration follows a hierarchical pattern:

1. Agent Definition: Each agent is defined in .cursor/agents/. For example, trip-planner.md defines its ability to collaborate with flight-researcher and hotel-researcher.
2. Skill Integration: Agents use skills defined in .cursor/skills/. Each skill maps to a specific action, such as executing a shell script.
3. Simulation Scripts: For this demo, flight and hotel searches are simulated using Bash scripts (research_flight.sh and research_hotel.sh) that return randomized results to illustrate the data flow.
4. Mermaid Generation: The trip-planner is instructed to generate a Mermaid diagram in the task logs to visualize the actual execution path taken during the task.

Example

Prompt to Cursor IDE

I submit the following prompt to Cursor IDE:

/trip-planner I want to go to San Francisco from NYC.

Please write a trip report in @reports/plan_to_san_francisco/trip_plan.md .

You have to comprehensively and precisely leave logs what and how you address the tasks with a diagram in Mermaid in @reports/plan_to_san_francisco/task_logs.md .

You have to show a diagram of what you address the task in the plan.

Result

• Cursor plan
• Trip report
• Task logs