# DeerFlow

DeerFlow는 샌드박스 도구, 메모리, 서브에이전트를 활용해 조사·코딩·창작을 수행하는 슈퍼 에이전트를 구축할 수 있는 LangChain 기반 오픈 소스 프레임워크입니다. 심층 연구, 분석, 콘텐츠 생성과 같은 장기 다단계 작업을 지원합니다.

- Findings24: <https://findings24.com/ko/tools/deerflow>
- 웹사이트: <https://deerflow.tech>
- 가격: Freemium
- 카테고리: AI 코드 어시스턴트

## DeerFlow란 무엇인가요?

DeerFlow is an open-source, LangChain-based framework for building super agents. The homepage describes it as a system that can research, code, and create with the help of sandboxes, memories, tools, skills, and subagents.

Its core purpose is to support long-running, multi-step work that may take minutes to hours. The site positions DeerFlow around deep research and task execution, with example workflows that include trend forecasting, data analysis, narrative-to-video generation, and explanatory content creation.

## DeerFlow(으)로 무엇을 할 수 있나요?

### Super-agent workflow

The homepage describes DeerFlow as a LangChain-based framework for building super agents that can handle tasks ranging from minutes to hours.

### Agent components

The product highlights sandboxes, memories, tools, skills, and subagents as the building blocks that let the agent handle different task levels.

### Context engineering

The site emphasizes long- and short-term memory so the agent can retain context across multi-step work.

### Planning and sub-tasking

It presents planning and sub-tasking as a built-in behavior, with the agent reasoning through complexity before executing sequentially or in parallel.

### Extensible tool stack

The homepage says DeerFlow supports extensible skills and tools, allowing built-in tools to be plugged in or swapped out.

### Multi-model support

The source also calls out multi-model support, listing Doubao, DeepSeek, OpenAI, Gemini, and others.

## 사용 사례

### Deep research and synthesis

Use DeerFlow to research a broad topic, synthesize sources, and produce a structured output such as a report or article. The homepage example about forecasting 2026 trends shows this research-first workflow.

### Content generation from source material

Use the agent to turn a source into a creative artifact, such as a webpage, video, or visual explanation. The homepage examples include generating a video from a novel scene and a comic strip explaining MOE.

### Exploratory data analysis

Use DeerFlow for data exploration when you want an agent to inspect a dataset, identify key factors, and summarize findings with insights. The Titanic dataset example points to this kind of analysis task.

### Long-running agent jobs

Use it when a task needs multiple steps, planning, and execution over a longer time horizon rather than a single prompt-response exchange. The site explicitly frames DeerFlow around long task running and sub-tasking.

### Tool-assisted build and iterate

Use it in workflows that benefit from a persistent sandbox and file access, such as writing code, saving artifacts, or iterating on outputs. The product highlights a mountable file system and a combined browser, shell, file, MCP, and VSCode environment.

## 자주 묻는 질문

### What is DeerFlow?

DeerFlow is presented as a LangChain-based framework for building super agents. The homepage frames it as an open-source system that can research, code, and create with help from sandboxes, memories, tools, skills, and subagents.

### What kinds of tasks is DeerFlow shown doing?

The source shows example workflows such as forecasting 2026 trends, generating a video from a novel scene, explaining MOE to a teenager, and analyzing the Titanic dataset. These examples suggest DeerFlow is meant for research-heavy, multi-step agent tasks.

### What does the sandbox setup emphasize?

The homepage describes an open-source AIO Sandbox and says it combines Browser, Shell, File, MCP, and VSCode Server in a single Docker container. It also highlights persistence, isolated execution, and mountable file system support.

### Is pricing available on the site?

The pricing URL currently returns a 404 page, so the source does not provide live pricing or plan details.

## 주요 정보

- **Category**: AI agent framework
- **Platform**: Web app and sandboxed agent workflow
- **Primary users**: Teams building or using research and task-running agents
- **License**: MIT license
- **Source domain**: deerflow.tech
- **Pricing page**: 404 / not available