Usage

Minimal cloud usage example

1. Navigate to https://agentgpt.reworkd.ai/.
2. In the Name field, type a short label for your agent, for example: ResearchAgent.
3. In the Goal field, describe the objective in plain English, for example:

   Research the top 5 open-source AI agent frameworks by GitHub stars and summarize their key differences.

4. Click Deploy Agent.
5. Watch the agent break the goal into tasks, execute each one, and display interim results in the task feed.
6. When the run completes (or the 5-run free limit is reached), copy the output from the task feed.

Pre-built templates

The cloud UI ships three ready-made templates you can launch with one click:

Template	What it does
ResearchGPT	Researches a topic and returns a structured report
TravelGPT	Plans an itinerary for a destination and travel window
StudyGPT	Builds a study plan for a subject or exam

Click a template, edit the goal if needed, then hit Deploy.

Minimal self-hosted API call (FastAPI backend)

After ./setup.sh completes, the FastAPI backend runs at http://localhost:8000. You can trigger an agent run via HTTP:

curl -X POST http://localhost:8000/api/agent/run \
  -H "Content-Type: application/json" \
  -d '{
    "goal": "List the 3 largest countries by area",
    "model_settings": {
      "model": "gpt-3.5-turbo",
      "max_tokens": 400,
      "temperature": 0.9
    }
  }'

The response is a stream of task events in JSON-lines format. Each event describes a task the agent created or completed.

Key concepts

Concept	Description
Goal	The high-level objective you give the agent
Task	A sub-step the agent creates to reach the goal
Run	One full execution cycle from goal to completion or limit
Model settings	Which GPT model to use, token budget, and temperature