Skip to main content

​
What you’ll build

In this quickstart, you’ll create a web research bot — an AI agent that takes a topic, searches the web, and delivers a structured summary. It takes about 5 minutes. By the end, you’ll understand the three building blocks of every MagOneAI bot:
  1. Agent — the AI persona that reasons and acts
  2. Workflow — the sequence of steps the agent follows
  3. Execution — a single run of the workflow with real input

​
Prerequisites

Before you start, make sure you have:
  • Access to a MagOneAI instance (Studio portal)
  • An LLM provider configured (GPT-4o, Claude, or any OpenAI-compatible model)
  • The Web Search tool available in your project
Don’t have Web Search configured yet? Check the Web Search setup guide — it takes about 2 minutes.

​
Step-by-step

1

Create a project

If you don’t have a project yet:
  1. Open MagOneAI Studio
  2. Click New Project
  3. Give it a name (e.g., “My First Bots”)
All your agents, workflows, and executions live inside a project.
2

Create a workflow (Use Case)

  1. Go to Use Cases in the left sidebar
  2. Click Create Use Case
  3. Name it “Web Research Bot”
  4. Open the Canvas (workflow builder)
A Use Case is your workflow — and agents are created inside it.
3

Add an Agent node and configure it

Your workflow needs just three nodes:
START --> Agent (Web Researcher) --> END
  1. The Start and End nodes are already on the canvas
  2. Drag an Agent node from the sidebar onto the canvas
  3. Connect: Start → Agent → End
  4. Click the Agent node to open its properties panel
  5. Configure the agent:
FieldValue
NameWeb Researcher
RoleResearch analyst
  1. Set the persona instructions:
You are a research analyst. Given a topic, search the web for current,
accurate information and produce a structured summary.

Your output must include:
- A 2-3 sentence overview
- 3-5 key facts or findings
- Sources you referenced

Be concise and factual. Cite your sources.
  1. Under Tools, add the Web Search tool
  2. Enable Can execute tools
  3. Click Save
That’s it — your agent and workflow are ready to run.
4

Run your bot

  1. Click Execute (or Run) on the canvas
  2. Enter your input:
{
  "query": "What are the latest developments in quantum computing?"
}
  1. Watch the execution progress:
    • Start node processes input
    • Agent node calls the LLM, which searches the web and synthesizes results
    • End node returns the final output
  2. View the result — a structured research summary with sources
5

Review the execution

After the run completes:
  1. Go to Executions in the sidebar
  2. Click on your execution
  3. Explore:
    • Activity logs — see input/output for each step
    • LLM usage — tokens used and cost
    • Duration — how long each activity took
Every execution is fully logged and auditable.

​
What just happened?

Here’s what MagOneAI did under the hood:

1. Temporal created a durable workflow

Your execution runs on Temporal — the same engine used by Uber and Stripe. If the server crashed mid-execution, it would resume exactly where it left off.
The LLM received your persona instructions, the user’s input, and the list of available tools. It decided to call the web search tool with relevant search queries.
The agent called the Web Search MCP server, which returned search results. The agent then synthesized these results into a structured summary — all within a single Agent node.
MagOneAI uses DSPy to ensure the agent produces structured, validated output — not just raw text. This makes outputs reliable and machine-readable.

​
Make it more powerful

Now that you have a working bot, here are quick ways to level it up:

Add conditional logic

Route to different agents based on the query type — questions go one way, research requests go another

Add human approval

Pause the workflow for human review before sending results to a customer

Run agents in parallel

Research multiple topics simultaneously — 3 agents running at the same time

Process files and data

Upload CSVs or PDFs and have agents analyze, summarize, or review them

​
Next: Build more bots

Ready to build something more advanced? Check out these quick bot recipes — each one takes under 10 minutes:

5 bots you can build today

Step-by-step recipes for support router, company research, competitor analysis, CSV analyzer, and document reviewer — no OAuth setup needed

Sales intelligence assistant

A full multi-agent workflow with parallel research — the most impressive demo