Agent · Infrastructure

Smolagents

Smolagents is Hugging Face's minimalist agent library. First-class support for code-writing agents where the model writes Python directly as its action space instead of JSON tool calls.

by Hugging Face·Fully open·Apache-2.0·Launched Dec 2024
Capability
n/a
no benchmarks
GitHub stars
26.5k
+134 · 7d
Harness
ReAct loop
Custom tools
Environments
1
Terminal

No public benchmark submissions · yet

Smolagents has not published applicable benchmark scores on SWE-bench Verified, GAIA, OSWorld or τ-bench. Frameworks are excluded from the capability leaderboard · they get ranked by adoption instead.

30-day star trajectory · velocity 57/100

25,80625,99226,17826,36326,5493/103/264/11STARS · 33 DAYS
Current
26,549
7d Δ
+134
30d Δ
+698

Brain · Harness · Tools · Environment

Brains
Which LLM powers it
n/a
Harness
The loop pattern
ReAct loop
Tools
How it acts on the world
Custom
Environments
Where it lives
Terminal
Not on the Mindshare watchlist yet. See tracked agents

Pulled from the agents dataset

What is Smolagents?

Smolagents is Hugging Face's minimalist agent library. First-class support for code-writing agents where the model writes Python directly as its action space instead of JSON tool calls.

Is Smolagents open-source?

Yes · Smolagents is fully open-source under the Apache-2.0 license. The harness, tools, and loop code are all public on GitHub at huggingface/smolagents.

How popular is Smolagents?

Smolagents has 26,549 stars on GitHub and gained 134 in the last 7 days across 200 contributors.

Brains used

No brains tracked

GitHub stats

Repohuggingface/smolagents
Stars26,549
7d Δ+134
30d Δ+698
Contributors200
Created2024-12-05
Last commit2026-04-02
How we score
Methodology

Composite capability · momentum formula · data sources.

Peer agents in the same category · sorted by capability