smolagents · Hugging Face
Hugging Face's minimalist (~1,000-line core, Apache 2.0) 'agents that think in code' harness where the model writes executable Python as its action format instead of JSON tool calls, model-agnostic via transformers/Ollama/LiteLLM, with optional sandboxed execution (E2B, Docker, WebAssembly), managed sub-agents, and Hub sharing of tools and agents.
Identity
| Homepage | https://huggingface.co/docs/smolagents/index |
|---|---|
| Repository | https://github.com/huggingface/smolagents |
| License | open-source |
| Domain | general |
| Form factor | sdk · cli |
Model support
| Model-agnostic | ✓ yes |
|---|---|
| Native models | — |
| Local models | ✓ yes |
| Pricing | byok |
Extensibility
| MCP client | ✓ yes |
|---|---|
| Agent Skills support | ○ no |
| Plugins / hooks | ✓ yes |
Execution & safety
| Sandboxing | optional |
|---|---|
| Autonomy spectrum | supervised → semi-autonomous → autonomous-local |
| Subagents | ✓ yes |
| Parallel sessions | ? unknown |
| Context management | none-observed |
Integration & evidence
| Headless / API | ✓ yes |
|---|---|
| GitHub integration | ? unknown |
| Sources |
✓ profile last verified 2026-07-08 · benchmark figures are vendor-reported unless a source states otherwise