Agents
tons-of-skills-marketplace
Model-agnostic marketplace of agent skills with a harness-free canonical layer.
★ 2.8k starsMIT
View guide →smolagents — Hugging Face's barebones agent library where agents think in code: a CodeAgent that writes its actions as Python, model-agnostic, Apache-2.0
smolagents is an open-source ai agents project by huggingface (huggingface/smolagents on GitHub). It has 29,927 GitHub stars and 3,106 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 6 October 2026.
Per the README: pip install "smolagents[toolkit]" for the package with the default tool set.
Create a model object — InferenceClientModel for Hub inference providers, or a local transformers/Ollama model, or another provider model.
Give a CodeAgent your tools (e.g. WebSearchTool) and the model, then call agent.run() with a task in plain language.
CodeAgent writes executable Python actions instead of JSON blobs — plans, loops and tool chaining come naturally and traces stay readable.
HF Inference Providers, local transformers/Ollama models and major hosted providers all plug into the same agent code.
A small, hackable codebase you can actually read end to end — the anti-framework agent library.
Tools are Python functions you can share on and pull from the Hugging Face Hub, alongside the bundled toolkit.
The README quick install with the default toolkit.
pip install "smolagents[toolkit]"The README example: a CodeAgent with web search on a Hub model.
from smolagents import CodeAgent, WebSearchTool, InferenceClientModel
model = InferenceClientModel()
agent = CodeAgent(tools=[WebSearchTool()], model=model)
agent.run("How many seconds would it take for a leopard at full speed to run through Pont des Arts?")Swap InferenceClientModel for a local transformers or Ollama model class from the docs — the agent code does not change.
https://huggingface.co/docs/smolagentsThe Hugging Face team (huggingface/smolagents on GitHub).
Yes — Apache-2.0 licensed. Model usage follows whatever provider or local model you plug in.
Code is more expressive than one-tool-per-step JSON: the agent can loop, branch, store intermediate results and compose tools in a single action, and the trace reads like a program you can debug. Execution is sandboxed per the library docs — still review what you let it run.
Against LangChain/LangGraph, smolagents is intentionally minimal — far less surface area, no graph DSL, quicker to learn and hack. Against AutoGen or CrewAI, it is a library for building your agent, not a multi-agent role-play system.
pip install "smolagents[toolkit]", then run the README CodeAgent example with InferenceClientModel and WebSearchTool.
Model-agnostic marketplace of agent skills with a harness-free canonical layer.
The agent that grows with you — a Python framework for adaptive AI agents.
A Claude Code-style coding agent that runs anywhere and works with almost any model provider
Secure sandboxes where AI agents run real code.
Python library for building multimodal language agents — VLM, video and mobile, fast to prototype and ship
Self-hosted ai agents: minimal ai coding agent (~1,000 lines of python) inspir.
The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if smolagents is not the right fit.
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
Free and open — grab the source, star the repo, and start creating.