New Course with E2B: Building Coding Agents with Tool Execution

Why It Matters

Enabling LLMs to write and safely execute code expands automation possibilities across software development and data analysis, giving businesses a competitive edge while mitigating security risks.

Summary

The video announces a new, jointly‑offered course with EDB titled “Building Coding Agents with Tool Execution,” taught by Teresa Tushkova and Francesco Zubigiri. It positions the curriculum as a hands‑on guide for developers who want to empower large language models (LLMs) to not only call predefined tools but also to write, run, and iterate on code as part of their problem‑solving workflow.

Key insights include the dramatic expansion of an agent’s functional space when it can generate and execute code on the fly. The presenter cites real‑world examples where agents discovered obscure Python libraries that solved tasks more elegantly than a human could have imagined, and where code generation enabled data acquisition, analysis, and visualization pipelines. However, the course also stresses the security pitfalls of unrestricted code execution, recommending sandboxed environments as a mandatory safeguard.

Notable details feature a step‑by‑step build‑from‑scratch of a simple Python‑executing agent that reads and writes files and interacts via chat. Learners then compare execution strategies—local, containerized, and MicroVM‑based sandboxes—before upgrading the agent to handle both Python and JavaScript in isolated H2B environments with strict resource limits. The capstone project transforms the agent into a data‑analyst bot that auto‑generates plots and ultimately a full‑stack Next.js web application, illustrating the end‑to‑end potential of code‑capable agents.

The implications are clear: developers who master these techniques can create far more autonomous AI assistants, accelerating software development cycles while maintaining security compliance. For enterprises, the ability to safely delegate coding tasks to agents promises cost reductions, faster time‑to‑market, and new avenues for AI‑driven innovation.

Original Description

Learn more: https://bit.ly/44tToAU

We’re excited to share a new course: Building Coding Agents with Tool Execution, created in collaboration with E2B, that teaches you how to build agents that can generate and execute code safely.

Many agent frameworks rely on predefined tools, but coding agents work differently: they generate code, run it, inspect the results, and plan their next step. This opens up far more capability, but it also introduces risk if the code runs on your local machine. That’s why the course focuses heavily on choosing and configuring safe execution environments.

In detail, you’ll learn:

- How coding agents reason, generate code, execute it, and iterate through feedback loops.

- The differences between local execution, containers, and isolated microVM sandboxes, and when each makes sense.

- How to run code safely in the cloud using E2B sandboxes, manage files, and control resources.

- How to build practical coding agents, including a data analyst agent that explores datasets and a full-stack agent that assembles a working Next.js web app.

Start building agents that handle more complex tasks while keeping your system protected.

Enroll now: https://bit.ly/44tToAU