{"slug":"smolagents","name":"smolagents","vendor":"Hugging Face","tagline":"Hugging Face's minimal Python library for agents that write their actions as code, with sandboxed execution and MCP tools","url":"https://agentboards.org/agent/smolagents","website":"https://huggingface.co/docs/smolagents/index","docs":"https://huggingface.co/docs/smolagents/index","repo":"https://github.com/huggingface/smolagents","category":"framework","execution":["local","sandbox"],"license":"Apache-2.0","open_source":true,"pricing":{"model":"byok","summary":"Free and Apache-2.0 licensed; you pay your model provider or run models locally","from_usd":0,"free_tier":true,"byok":true},"models":{"backbone":["Hugging Face Inference providers","OpenAI","Anthropic","any LiteLLM model","Transformers","Ollama"],"bring_your_own_model":true,"local_models":true},"protocols":{"mcp_client":true,"mcp_server":false,"openapi":false},"capabilities":{"terminal_exec":false,"browser":true,"multi_file_edit":false,"git_ops":false,"docker_sandbox":true,"multi_agent":true,"headless_ci":true},"context_window":null,"platforms":["macos","linux","windows"],"install":[{"label":"pip","command":"pip install 'smolagents[toolkit]'"}],"benchmarks":[],"tags":["framework","python","code-agents","sandbox","e2b","docker","mcp","hugging-face"],"github_stars":29638,"metrics":{"stars":29638,"pushed_at":"2026-09-30T05:07:25Z","archived":false,"open_issues":870,"latest_version":"v1.26.0","latest_version_at":"2026-05-29T05:09:47Z","latest_version_url":"https://github.com/huggingface/smolagents/releases/tag/v1.26.0","version_source":"github","checked_at":"2026-10-02T05:02:38.750Z"},"sources":{"pricing":"https://github.com/huggingface/smolagents","license":"https://github.com/huggingface/smolagents","install":"https://huggingface.co/docs/smolagents/index","models":"https://huggingface.co/docs/smolagents/index","protocols":"https://huggingface.co/docs/smolagents/index","capabilities":"https://huggingface.co/docs/smolagents/reference/default_tools"},"verified_at":"2026-09-03","adoption":{"score":6.8,"signals":{"pypi_weekly":101432,"github_stars":29506,"hn_mentions_1y":2},"measured_at":"2026-09-28"},"scores":{"panel":6.7,"adoption":6.8,"spread":1.8,"facts":10,"board":7.2,"dimensions":{"reliability":6.2,"usefulness":6.3,"cost":8,"longevity":6.3},"per_persona":{"amigo":7,"critico":6.3,"profesor":6.3,"inversora":6.5,"jefa":6.3,"hacker":8},"community":null},"rank":null,"category_rank":16,"panel_reviews":[{"persona":"juez","persona_name":"El Juez","ai_generated":true,"verdict":"The split is under two points and it is about one constructor argument: El Hacker owns the code, El Crítico notes the sandbox is opt-in.","scores":null,"evidence":[]},{"persona":"amigo","persona_name":"El Amigo","ai_generated":true,"verdict":"Pick smolagents if you want to read every line of your agent framework in an hour and run it on a local model; pick LangGraph when the job needs persistence and resumption.","scores":{"reliability":6,"usefulness":7,"cost":9,"longevity":6},"evidence":["https://huggingface.co/docs/smolagents/index","https://github.com/huggingface/smolagents"]},{"persona":"critico","persona_name":"El Crítico","ai_generated":true,"verdict":"CodeAgent executes model-written Python, and the sandbox is opt-in, so the default path runs generated code on the machine that called it.","scores":{"reliability":5,"usefulness":6,"cost":8,"longevity":6},"evidence":["https://huggingface.co/docs/smolagents/index"]},{"persona":"profesor","persona_name":"El Profesor","ai_generated":true,"verdict":"Actions are Python code rather than JSON tool calls, justified by composability: nesting, loops and conditionals in one step; a ToolCallingAgent covers the conventional path.","scores":{"reliability":6,"usefulness":6,"cost":7,"longevity":6},"evidence":["https://huggingface.co/docs/smolagents/index","https://github.com/huggingface/smolagents"]},{"persona":"inversora","persona_name":"La Inversora","ai_generated":true,"verdict":"Hugging Face gives the library away because the default model class calls its Inference providers; the framework is a funnel to a marketplace, which is a durable reason to keep maintaining it.","scores":{"reliability":7,"usefulness":6,"cost":6,"longevity":7},"evidence":["https://huggingface.co/docs/smolagents/index","https://github.com/huggingface/smolagents"]},{"persona":"jefa","persona_name":"La Jefa","ai_generated":true,"verdict":"As a dependency it costs nothing, has no support contract, and ships Hub integration that lets an engineer publish an agent as a public Space, which is the governance conversation to have first.","scores":{"reliability":5,"usefulness":6,"cost":8,"longevity":6},"evidence":["https://huggingface.co/docs/smolagents/index","https://github.com/huggingface/smolagents"]},{"persona":"hacker","persona_name":"El Hacker","ai_generated":true,"verdict":"Apache-2.0, TransformersModel loads weights on my own GPU, Ollama works, ToolCollection.from_mcp pulls any MCP server, and Tool.from_langchain steals from the other ecosystem.","scores":{"reliability":8,"usefulness":7,"cost":10,"longevity":7},"evidence":["https://github.com/huggingface/smolagents","https://huggingface.co/docs/smolagents/index"]},{"persona":"comediante","persona_name":"El Comediante","ai_generated":true,"verdict":"Ships a command-line tool called webagent, so the thousand lines also browse.","scores":null,"evidence":[]}]}