NeuronetworksAI Ed

Home / Projects / AgentFlow AI

Year 3 · AI/ML Engineering

AgentFlowAI

Build an AI agent that can plan complex tasks, break them down into steps and use tools such as web search, code execution and file operations to complete real-world tasks.

  • Python
  • Agents
  • Tool Calling
  • APIs

Illustrative demo interface. Based on the roadmap project: Tool-using AI agent.

AgentFlow AI▸ Run
New task
  • History
  • Tools
  • Settings

Find the latest research on RAG, create a short summary and save it to a markdown file.

  1. Planning taskBreaking down the request into steps
  2. Searching the webUsing web_search tool
  3. Analysing resultsReading and summarising key points
  4. Creating fileSaving output to rag_summary.md
Task completedSummary of the latest RAG research saved to rag_summary.md.
Open file

Tool calls

  • web_searchQuery: latest RAG research● Completed
  • pythonAnalyse and summarise● Completed
  • file_writeSave to rag_summary.md● Completed
4 / 4 steps donePlan executed
3 tools calledSearch, code, files

Project overview

What you build.

You will build an AI agent that can understand a user's request, plan the steps needed, choose the right tools and complete the task with verifiable results.

User requestFind information, run analysis, create a file…
AI agent
  • Understand
  • Plan steps
  • Choose tools
  • Execute
  • Verify results
Web searchGet latest information
Code executionRun Python code
File operationsRead and write files
Final responseTask completed with output you can verify.

Step by step

How the agent
works.

The agent follows a clear process from understanding your request to delivering a final result, using the right tools at each step.

  1. 01Interpret requestUnderstand what you are asking for.
  2. 02PlanBreak the task into smaller steps.
  3. 03Choose toolSelect the most appropriate tool for each step.
  4. 04ExecuteRun the tool and observe the results.
  5. 05Verify resultCheck the output and handle errors if needed.
  6. 06RespondProvide the final answer with references or files.

Project walkthrough

Inside the workflow.

Work in a development environment where you can see the agent's reasoning steps, tool calls and results at each step.

  • View the agent's plan and execution steps
  • Inspect tool calls and inputs
  • See intermediate results and final output
  • Try different requests and test edge cases
AgentFlow AIIllustrative demo interface
New task
  • Projects
  • Playground
  • Tools
  • Logs

Execution flow

  1. User requestFind the latest research on RAG, create a short summary and save it to a markdown file.
  2. Plan6 steps
  3. web_searchSearch for recent RAG research
  4. pythonAnalyse and summarise content
  5. file_writeSave to rag_summary.md
  6. Final responseTask completed successfully
TraceLogsOutput
{
  "step": "web_search",
  "query": "latest RAG research",
  "status": "completed",
  "results": [
    {
      "title": "Retrieval-augmented…",
      "url": "https://…"
    }
  ]
}

Result preview

rag_summary.mdMarkdown
View file

Architecture

Tools and
architecture.

Learn how an agent is built using a language model, a tool registry and a controlled execution environment.

User request
Agent orchestrator(LLM + tool registry)
Validated resultSafe and useful output
Web searchSearch the internet
Code executionRun Python code
File operationsRead and write files
Other toolsAPIs and custom tools

Safety by design

Safety and
control.

Learn how to build agents that are useful, reliable and safe, with clear permissions and human oversight where needed.

  • Permission controlLimit what tools can do and which resources they can access.
  • Approval for sensitive actionsRequire confirmation before critical operations such as file changes.
  • Execution logsKeep a record of tool calls, inputs and outputs for transparency.
  • Result validationCheck and verify outputs to reduce errors and hallucinations.

Tech stack

Tools and
technologies.

Work with modern tools and frameworks used to build AI agents.

  • PythonCore language
  • LLM APIsProvider APIs
  • Function callingTool integration
  • Web searchSearch APIs
  • Sandboxed executionRun code safely
  • File handlingRead and write files

Skills you practise

Key skills
you will gain.

This project helps you build practical skills in agent design, tool use and real-world problem solving.

  • Task decompositionBreak complex requests into steps
  • Result verificationCheck and validate outputs
  • Tool calling and integrationUse APIs and external tools
  • Safe executionApply permissions and controls
  • Error handlingHandle tool failures and edge cases
  • Debugging and iterationTest, improve and refine the agent

Program context

Part of the AI/ML
Engineering program.

This project is part of Year 3, where you build advanced topics and work on real-world applications with a portfolio of projects.

AI/ML Engineering

Year 3 · Specialisation

From Python basics to machine learning, deep learning, generative AI and MLOps, with a portfolio of projects.

View program
  • Advanced AI and LLMs
  • RAG and agent development
  • MLOps and deployment
  • Portfolio of projects

Your next step

Ready to explore more projects?

Discover more hands-on projects from the AI/ML Engineering program and build a portfolio you can show.