smolagents Documentation & Guided Tour
by Hugging Face
A minimalist agent library whose docs are a fast, code-first way to grok how code-writing agents work.
Overview
smolagents is a deliberately small library — the whole thing is a few thousand lines — which makes its documentation an unusually clear window into how modern agents work. Its signature idea is 'code agents': instead of emitting JSON tool calls, the agent writes Python that calls your tools, which is often more expressive and reliable. The guided tour and how-to guides get you from zero to a working tool-using agent quickly.
At a Glance
- Topic
- Agentic
- Level
- Intermediate
- Format
- Documentation
- Cost
- Free
- Duration
- Self-paced
- Provider
- Hugging Face
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓The 'code agent' paradigm (agents that write Python)
- ✓Defining tools and secure execution sandboxes
- ✓Multi-step and multi-agent orchestration in a tiny footprint
- ✓Connecting any LLM backend
Highlights
- •Tiny, readable codebase — great for understanding internals
- •Code-as-action approach
- •Works with local or hosted models
Who It’s For
Best For
- ✓Developers who prefer minimal, transparent frameworks
Prerequisites
- •Intermediate Python
FAQ
What is smolagents Documentation & Guided Tour?
The official docs for smolagents, Hugging Face's tiny agent framework built around agents that write and run Python code as their action space.
Is smolagents Documentation & Guided Tour free?
smolagents Documentation & Guided Tour is free to access.
What level is smolagents Documentation & Guided Tour for?
smolagents Documentation & Guided Tour is aimed at a intermediate audience. Recommended background: Intermediate Python.
How long does smolagents Documentation & Guided Tour take?
Expect roughly Self-paced. Most learners work through it at their own pace.
What will I learn from smolagents Documentation & Guided Tour?
You'll learn: The 'code agent' paradigm (agents that write Python); Defining tools and secure execution sandboxes; Multi-step and multi-agent orchestration in a tiny footprint; Connecting any LLM backend.