DeerFlow is a community-driven framework designed for deep research, integrating language models with specialized tools for tasks like web search, crawling, and Python code execution. It offers a modular multi-agent system architecture for automated research, supporting various search engines, crawling tools, and private knowledge bases.

deer-flow

Active·★ 83.5k·MIT·Updated 2026-10-08

★ Most Popular★ Trending★ Essential

DeerFlow is ByteDance's open-source "super agent harness": a runtime that gives an AI agent a filesystem, sandboxed code execution, long-term memory, skills and sub-agents so it can carry out tasks that take minutes to hours. Version 2.0 is a ground-up rewrite built on LangGraph and LangChain; the original Deep Research framework is still maintained on the 1.x branch.

It suits developers and teams who want a ready-made agent runtime rather than wiring one together from libraries. Typical tasks include in-depth research reports, building data pipelines, generating slide decks and dashboards, and automating content workflows. Tasks can be sent from the web workspace or from messaging apps such as Telegram, Slack, Feishu/Lark, WeChat, WeCom, DingTalk and QQ, and it can retrieve from private knowledge bases through RAGFlow.

Getting started: Docker is the recommended path (Docker Compose v2.24+), with `make docker-init` and `make docker-start` after configuring models. For local evaluation plan for at least 4 vCPU, 8 GB RAM and 20 GB of free disk; a long-running shared server should start at 8 vCPU and 16 GB RAM. Linux with Docker is the recommended target for persistent deployments.

#LLM#Multi-Agent#Research Automation#Web Search#RAG#Web Browsing#Coding

Features

01Multi-model LLM integration via LiteLLM with OpenAI-compatible API

02Comprehensive web search and crawling using diverse engines like InfoQuest and Tavily

03RAG integration with multiple vector databases including Qdrant and RAGFlow

04Human-in-the-loop collaboration with intelligent clarification and interactive plan modification

05Automated content creation for podcasts and presentations with AI-powered script generation

Why choose it

+Batteries included: sandbox, filesystem, memory, skills, sub-agents and scheduled tasks come out of the box instead of being assembled by hand.

+Handles long-running, multi-step tasks, with context compaction and sub-agent delegation designed for them.

+Reachable from many messaging apps (including Feishu, WeChat, WeCom and DingTalk) without needing a public IP.

+MIT licensed, with tracing integrations (LangSmith, Langfuse) and support for multiple model providers.

Trade-offs

!Resource-hungry: 2 vCPU / 4 GB is usually not enough, and shared servers need 8 vCPU / 16 GB or more.

!It has high-privilege capabilities (system commands, file operations) and is designed for a trusted local environment; the project explicitly warns against exposing it to the internet without proper access control.

!Version 2.0 shares no code with 1.x, so tutorials and integrations written for the original Deep Research framework may not apply.

!A large, fast-moving codebase with many configuration options; expect time spent on setup and tuning.

Compatibility

LiteLLM

Native Integration

Verified via docs

OpenAI-compatible APIs

API Support

Verified via docs

InfoQuest

Recommended Search

Verified via docs

Tavily

Default Search

Verified via docs

Jina

Default Crawler

Verified via docs

RAGFlow

RAG Provider

Verified via docs

Quick start

$ uv sync

Use cases

↳Conducting deep research and generating comprehensive reports with multimedia content

↳Automated creation of podcasts, articles, and presentations from research findings

↳Intelligent information retrieval and analysis for complex queries and trending topics

How it compares

≈vs GPT Researcher: GPT Researcher focuses on producing cited research reports; DeerFlow 2.0 is a general agent runtime in which research is only one of many task types.

≈vs OpenClaw: OpenClaw is a personal assistant that lives in your chat apps; DeerFlow is aimed at heavier work tasks with sandboxes, sub-agents and project workspaces.

≈vs building on LangGraph directly: DeerFlow is built on LangGraph but ships the memory, sandbox and skills layers that you would otherwise implement yourself.

Alternatives

Microsoft AutoGen★ 61.3k

A framework that enables the development of LLM applications using multiple agents that can converse with each other to solve tasks.

vs →

CrewAI★ 59.5k

Framework for orchestrating role-playing, autonomous AI agents. By working together, your Crew can tackle complex tasks.

vs →

ragflow★ 91.9k

RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context layer for LLMs

Context7★ 62.8k

MCP Server that provides up-to-date code documentation for LLMs and AI code editors.

GitHub MCP Server★ 33.5k

GitHub's official MCP Server. Allows AI agents to interact directly with your GitHub repositories (read files, search code, issues).

n8n★ 206.7k

Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations.

Brave Search MCP★ 1.5k

Allow your AI Agent to search the real-time internet using Brave Search API. Essential for getting up-to-date information.

ToolJet★ 41.1k

ToolJet is the open-source foundation of ToolJet AI - the AI-native platform for building internal tools, dashboard, business applications, workflows and AI agents 🚀

See all alternatives →

Comments

No comments yet. Be the first!

deer-flow

Active·★ 83.5k·MIT·Updated 2026-10-08

★ Most Popular★ Trending★ Essential

#LLM#Multi-Agent#Research Automation#Web Search#RAG#Web Browsing#Coding

Features

01Multi-model LLM integration via LiteLLM with OpenAI-compatible API

02Comprehensive web search and crawling using diverse engines like InfoQuest and Tavily

03RAG integration with multiple vector databases including Qdrant and RAGFlow

04Human-in-the-loop collaboration with intelligent clarification and interactive plan modification

05Automated content creation for podcasts and presentations with AI-powered script generation

Why choose it

+Batteries included: sandbox, filesystem, memory, skills, sub-agents and scheduled tasks come out of the box instead of being assembled by hand.

+Handles long-running, multi-step tasks, with context compaction and sub-agent delegation designed for them.

+Reachable from many messaging apps (including Feishu, WeChat, WeCom and DingTalk) without needing a public IP.

+MIT licensed, with tracing integrations (LangSmith, Langfuse) and support for multiple model providers.

Trade-offs

!Resource-hungry: 2 vCPU / 4 GB is usually not enough, and shared servers need 8 vCPU / 16 GB or more.

!Version 2.0 shares no code with 1.x, so tutorials and integrations written for the original Deep Research framework may not apply.

!A large, fast-moving codebase with many configuration options; expect time spent on setup and tuning.

Compatibility

LiteLLM

Native Integration

Verified via docs

OpenAI-compatible APIs

API Support

Verified via docs

InfoQuest

Recommended Search

Verified via docs

Tavily

Default Search

Verified via docs

Jina

Default Crawler

Verified via docs

RAGFlow

RAG Provider

Verified via docs

Quick start

$ uv sync

Use cases

↳Conducting deep research and generating comprehensive reports with multimedia content

↳Automated creation of podcasts, articles, and presentations from research findings

↳Intelligent information retrieval and analysis for complex queries and trending topics

How it compares

≈vs GPT Researcher: GPT Researcher focuses on producing cited research reports; DeerFlow 2.0 is a general agent runtime in which research is only one of many task types.

≈vs OpenClaw: OpenClaw is a personal assistant that lives in your chat apps; DeerFlow is aimed at heavier work tasks with sandboxes, sub-agents and project workspaces.

≈vs building on LangGraph directly: DeerFlow is built on LangGraph but ships the memory, sandbox and skills layers that you would otherwise implement yourself.