# Deep Research Smolagents

> Run Hugging Face Smolagents as a code-as-action deep research agent with optional web search and model backends. Use when a benchmark needs explicit intermediate tool actions, code execution, or an open-source agent framework.

- Skill: `closedloop-technologies/deep-research-smolagents` (Agent Skill, multi-file: 5 files)
- Install (CLI): `npx skillmds@latest add closedloop-technologies/deep-research-smolagents`
- Raw SKILL.md: https://api.skillmd.com/api/skills/closedloop-technologies/deep-research-smolagents/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: closedloop-technologies (https://skillmd.com/u/closedloop-technologies)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/closedloop-technologies/deep-research-smolagents

---


# Smolagents

Use this skill when the research task benefits from tool use, code execution,
and explicit intermediate actions.

## Environment

```bash
op run --env-file .env.climbhill -- python .agents/skills/deep-research-smolagents/scripts/agent.py \
  --task "Find current deep research APIs and summarize cost controls" \
  --model openai \
  --model-id gpt-4o-mini \
  --web-search
```

Common 1Password-backed variables:

- `HF_TOKEN`
- `OPENAI_API_KEY`
- `ANTHROPIC_API_KEY`

## Cost Controls

- Use economical model IDs for benchmark tasks.
- Keep `max_steps` low in the script configuration.
- Enable web search only when the prompt requires current sources.

## Source Link

- https://github.com/huggingface/smolagents