deer-flow vs vibe-trading
Two of the top projects in Agent platforms, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
deer-flow
Agent harness that runs sub-agents, memory and sandboxed skills for long tasks
vibe-trading
Finance research agent for backtests, market data and broker connectors
Verdict
Both score 77; deer-flow ranks higher on stars, 84,230 to 36,101. Of the 5 score parts we measure for both, deer-flow leads on 2, vibe-trading on 2 and 1 is even.
- Adoptiondeer-flow leads by 21 points, 91 to 70: a wide gap.
- FreshnessEven at 100.
- Maintenancevibe-trading leads by 23 points, 99 to 76: a wide gap.
- Easy to runvibe-trading leads by 17 points, 67 to 50: a wide gap.
- Agent-readydeer-flow leads by 70 points, 70 to 0: a wide gap.
Pick deer-flow if you are
Engineers self-hosting a multi-agent harness for research and long-running tasks
- Interactive `make setup` wizard generates config.yaml and .env in about 2 minutes
- Web UI model registration for OpenAI, Anthropic, DeepSeek, Qwen, Doubao, Kimi, Zhipu, MiniMax
Pick vibe-trading if you are
Quant researchers and traders who want an LLM agent for backtesting
- Grounding gate checks each reported figure against session tool evidence and cuts failures
- Backtest run cards cite metrics only where saved CSV output agrees
| What we compare | deer-flow | vibe-trading |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 91, widely used | 70, popular |
| Freshness | 100, active | 100, active |
| Maintenance | 76, fair | 99, healthy |
| Easy to run | 50, easy | 67, easy |
| Agent-ready | 70, partly | 0, none |
| Facts from GitHub and the README | ||
| Stars | 84.2k | 36.1k |
| License | MIT (permissive) | MIT (permissive) |
| Last commit | Oct 2026 | Oct 2026 |
| Last release | Sep 2026 | Sep 2026 |
| Language | Python | Python |
| Docker | Yes | Yes |
| GPU | Not needed | Not needed |
| arm64 or Apple Silicon | Mentioned | Mentioned |
deer-flow
DeerFlow 2.0 is a ByteDance agent harness where a lead agent delegates to sub-agents, backed by long-term memory, a sandboxed file system and extensible skills. It includes a web UI, a Gateway, MCP support, IM channels, scheduled tasks, a terminal UI and an embedded Python client. It is a ground-up rewrite that shares no code with v1.
Who it is for: Engineers self-hosting a multi-agent harness for research and long-running tasks
Strengths
- Interactive `make setup` wizard generates config.yaml and .env in about 2 minutes
- Web UI model registration for OpenAI, Anthropic, DeepSeek, Qwen, Doubao, Kimi, Zhipu, MiniMax
- Tracing via LangSmith, Langfuse and Monocle, plus `make doctor` and `make support-bundle`
- Many search and fetch backends: Brave, Tavily, Jina, Firecrawl, Crawl4AI, InfoQuest
Weaknesses
- README warns improper deployment carries security risks; Gateway admin equals code execution
- Local development needs Python 3.12+ and Node.js 22+
- README is dense with config detail; RAM and hardware needs are not stated
- v2 shares no code with v1; v1 users must stay on the 1.x branch
- no GPU
- Docker + Compose
- Needs LLM provider API key, Python 3.12+, Node.js 22+
- Models: OpenAI, Anthropic, DeepSeek, Qwen, Doubao
vibe-trading
Vibe-Trading is a Python agent (FastAPI backend, React 19 web UI, CLI, MCP) for trading research. It pulls market data across many exchanges, runs backtests and factor analysis, and checks figures in generated reports against tool evidence. It also connects to 18 brokers and sends scheduled reports through channels such as Email, Slack, Discord and Feishu.
Who it is for: Quant researchers and traders who want an LLM agent for backtesting
Strengths
- Grounding gate checks each reported figure against session tool evidence and cuts failures
- Backtest run cards cite metrics only where saved CSV output agrees
- 18 broker connectors, several read-only, with paper/live permissions documented
- Active: v0.1.16 shipped 116 merged PRs, with near-daily fixes afterward
Weaknesses
- Narrow domain: finance research only, not a general agent builder
- Some broker features are unavailable, e.g. Robinhood options execution
- Data quality depends on public sources such as Stooq, which can block requests
- Frequent backtest-correctness fixes in the changelog; verify results independently
- no GPU
- Docker + Compose
- Needs LLM provider (OpenRouter is the default), Market data sources, Broker APIs (optional)
- Models: OpenRouter, Codex, Copilot, custom model endpoints
More in Agent platforms
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.