Logo
Deploy Now

SaaS Alternative

OpenAI Assistants

Categories

Self-Hosted Developer Tools Automation LLM AI Agent Multi-Agent Deep Research

Stars

80,493

Forks

11,044

Watchers

337

Developer links

Last Updated: 08-21-2026 09:42 PM (UTC)

DeerFlow

DeerFlow 2.0 is ByteDance's ground-up rewrite that transforms what began as a Deep Research framework into a batteries-included super agent harness handling tasks lasting minutes to hours autonomously. Built on LangGraph and LangChain, the runtime orchestrates a lead agent that plans, decomposes, and delegates to sub-agents executing in isolated Docker or Kubernetes sandboxes with persistent filesystem access, while an extensible skills system lets developers add capabilities as Python functions or MCP servers with OAuth token flows. The harness ships with long-term memory using persistent event loops with per-agent isolation, scheduled task execution via cron, context engineering with manual compaction, and a web UI at localhost:2026 for interactive sessions. Model support spans OpenAI GPT-4o/GPT-5, Anthropic Claude via OAuth, Google Gemini, DeepSeek, Qwen via vLLM, and OpenRouter-compatible gateways with per-model pricing configuration. IM channel integration connects Telegram, Slack, Feishu/Lark, Discord, WeChat, WeCom, DingTalk, and Buzz without requiring a public IP — all using long-polling or WebSocket transports. Production deployment uses Docker Compose with nginx reverse proxy, PostgreSQL or SQLite persistence, Redis stream bridges for multi-worker scaling, and lease-based run ownership with automatic orphan recovery. The terminal workbench TUI enables headless operation. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

DeerFlow
DeerFlow
DeerFlow

Benefits

  • Sandboxed Long-Horizon Execution
  • Agents execute code, manage files, and install dependencies inside Docker or Kubernetes sandboxes with persistent filesystem access, enabling multi-step tasks running minutes to hours autonomously.
  • Extensible Skills and MCP Support
  • Add capabilities as Python functions or MCP servers with OAuth flows, per-tool timeouts, and namespace prefixing, while built-in skills handle research, podcasts, and code generation.
  • Multi-Model Provider Flexibility
  • Configure OpenAI, Anthropic Claude Code OAuth, Google Gemini, DeepSeek, Qwen via vLLM, and OpenRouter gateways with per-model pricing, reasoning toggles, and Responses API support.
  • IM Channel Integration Without Public IP
  • Receive tasks from Telegram, Slack, Feishu/Lark, Discord, WeChat, WeCom, and DingTalk via long-polling or WebSocket transports without requiring callback URLs or public endpoints.

Features

  • Sub-Agent Orchestration
  • Lead agent decomposes tasks and spawns sub-agents that execute in parallel with independent sandboxes, token tracking, and Langfuse trace attribution back to parent threads.
  • Persistent Memory System
  • Long-term memory runs on a persistent event loop with per-agent isolation, queued updates, and SOUL.md files that agents can self-edit to update their own configuration.
  • Docker/Kubernetes Sandboxes
  • Code execution happens in isolated containers with configurable modes — local, Docker, or Kubernetes via provisioner — with lease-based ownership and orphan recovery.
  • Production Multi-Worker Scaling
  • Redis stream bridges, lease heartbeats, and atomic run ownership enable multi-worker Gateway deployments with automatic dead-worker reconciliation and SSE gap recovery.
  • Scheduled Task Execution
  • Cron-based task scheduling runs agent workflows unattended with configurable recurrence, enabling automated research, monitoring, and content generation pipelines.