Build coding agents that write and execute Python code, manage files, and handle errors through iterative feedback loops.
Intermediate1h16m
Building Coding Agents with Tool Execution
Instructors: Tereza Tizkova, Francesco Zuppichini
Earn an accomplishment with PRO · Updated Jan 10, 2026
- Intermediate
- 1h16m
- 9 Video Lessons
- 5 Code Examples
- 1 Graded Assignment PRO
- Earn an accomplishment with PRO
- Updated Jan 10, 2026
- Instructors: Tereza Tizkova, Francesco Zuppichini
E2B- Learn more aboutMembership PRO Plan
What you'll learn
Run agent code safely in isolated E2B cloud sandboxes and understand the tradeoffs between local, containerized, and cloud execution environments.
Create specialized agents: a data analyst that explores datasets and visualizes insights, and a full-stack agent that builds complete web applications.
Course recap
PROThis course, built in partnership with E2B, teaches how to build coding agents from scratch that generate and execute code safely in cloud sandbox environments. Unlike agents with predefined tool sets, coding agents can write arbitrary Python code, giving them access to every built-in function and installable package. The course progresses from a simple agent with code execution and file system tools to a full-stack agent that builds complete Next.js web applications inside sandboxes, with runtime context summarization to manage long conversations.
Concept map
Key concepts
- AI AgentsCoding agents defined as LLMs calling tools in a loop with context, writing and executing arbitrary code
- Sandbox environmentsMicroVM-based isolated environments (like AWS Firecracker) that provide kernel-level isolation for untrusted code execution
- Context engineeringKeeping context small while providing relevant information; includes tool output truncation, pagination, and runtime summarization
- Tool executionStructured JSON schemas that describe tools, with execute functions that run code and return results
- Runtime summarizationCompressing 70% of oldest messages using an LLM to maintain context quality in long sessions
Lesson highlights
- 1.Introduction — What coding agents are and why sandbox execution matters
- 2.Inside a Coding Agent — LLM selection, context management, tool design, and execution environments
- 3.Your First Coding Agent — Building an agent with code execution, file read/write tools, and multi-step loops
- 4.Tool Execution Environments — Local vs Docker vs sandbox (microVM) trade-offs
- 5.Running Code in the Cloud — E2B sandbox APIs for creating, managing, and interacting with remote environments
- 6.Data Analyst Agent — CSV exploration and visualization agent with Gradio UI
- 7.Full Stack Agent — Multi-file editing agent that builds Next.js web apps with runtime summarization
- 8.Hands-On Project Introduction — Building and deploying your own sandbox agent
- 9.Conclusion — Course recap
About this course
Join our new short course, Building Coding Agents with Tool Execution! Learn from Tereza Tizkova, Growth, and Francesco Zuppichini, Machine Learning Engineer at E2B.
In this course, you’ll build agents that write and execute code to accomplish tasks, going far beyond predefined function calls. Instead of limiting your agent to a fixed set of tools, you’ll let it access Python’s entire ecosystem, and write multi-step code sequences. Plus, you’ll learn to run this generated code safely in sandboxed cloud environments, protecting your systems from potentially harmful operations.
In detail, you’ll:
- Understand coding agents and build your first one – Learn what makes coding agents different from other AI agents, then build an agent that executes Python code, manages files, and handles errors through feedback loops.
- Choose and configure safe execution environments – Compare local execution, containers, and sandboxed microVMs, then run your agent’s code in secure E2B cloud environments with built-in isolation and resource controls.
- Build a data analyst agent – Create an agent that explores CSV datasets, generates summaries and visualizations with Pandas, and answers questions through a Gradio chat interface.
- Deploy a full-stack coding agent – Build an agent that edits multiple files and generates complete Next.js web applications in a sandbox, while managing long contexts through runtime summarization.
Start building agents that can code their way through complex tasks safely and efficiently.
Who should join?
AI builders familiar with Python and LLMs who want to build agents that can write and execute code. Basic understanding of AI agents is helpful but not required.
Course Outline
9 Lessons・5 Code Examples- IntroductionVideo・3m
- Inside a Coding AgentVideo・8m
- Your First Coding AgentVideo with Code Example・10m
- Tool Execution EnvironmentsVideo・5m
- Running Code in the CloudVideo with Code Example・7m
- Data Analyst AgentVideo with Code Example・4m
- Full Stack AgentVideo with Code Example・8m
- ConclusionVideo・1m
- Hands-On Project Introductory VideoVideo・6m
- Hands-On ProjectCode Example・10m
- Quiz
Graded・Quiz
・12m
Elevate your learning experience with Pro
Upgrade to Pro and gain unlimited accomplishments on your resume
Instructors
Additional learning features, such as quizzes and projects, are included with DeepLearning.AI Pro. Explore it today
Want to learn more about Generative AI?
Keep learning with updates on curated AI news, courses, and events, as well as Andrew’s thoughts from DeepLearning.AI!