# FiClaw

> FiClaw 是 AI Native 量化策略研发工作站：用一句话描述策略想法，AI 自动完成代码生成、真实回测、诊断迭代和参数优化。它内置金融 Agent 桌面端，服务投研、量化与风控团队从研究到执行的完整工作流。

## 产品

- [产品能力](https://ficlaw.ai/products): FiClaw 桌面端、内置金融 Agent、账号积分与云端模型路由的整体能力。
- [工作原理](https://ficlaw.ai/how-it-works): 从策略想法到回测验证的自动化研发链路。
- [量化能力](https://ficlaw.ai/quant-capabilities): 因子研究、策略生成、真实回测与参数优化。
- [Agents](https://ficlaw.ai/agents): 金融 Agent 如何进入研究、风控与执行流程并承接角色分工。
- [集成与接口](https://ficlaw.ai/integrations): 数据、模型与量化能力的对接方式。
- [A 股数据覆盖与回测边界](https://ficlaw.ai/data/a-share-backtest-coverage): QuantAPI、沪深 300、10 年以上回测区间、复权、停牌、费用、滑点和幸存者偏差说明。
- [下载](https://ficlaw.ai/download): FiClaw Desktop（macOS Apple Silicon / Windows x64）。
- [快速开始](https://ficlaw.ai/quick-start): 首次使用 FiClaw 的上手路径。
- [模型与积分](https://ficlaw.ai/pricing): 云端模型路由与账号积分说明。
- [示例回测报告](https://ficlaw.ai/examples): 查看 prompt、策略摘要、指标、诊断和下一步优化建议。
- [方法论](https://ficlaw.ai/methodology): FiClaw 如何组织 spec、代码生成、真实回测、参数优化和人工复核边界。
- [A 股常见选股因子库](https://ficlaw.ai/factors): 估值、质量、成长、动量、低波动和流动性因子的定义、公式、A 股回测口径与常见陷阱。
- [估值因子](https://ficlaw.ai/factors/value): PE、PB、PS 与股息率的计算、行业中性化和验证方法。
- [质量因子](https://ficlaw.ai/factors/quality): ROE、毛利率、现金流与杠杆的组合研究方法。
- [成长因子](https://ficlaw.ai/factors/growth): 营收与利润增长、基数效应和披露滞后处理。
- [动量因子](https://ficlaw.ai/factors/momentum): 20、60、120 日动量、复权与涨跌停成交约束。
- [低波动因子](https://ficlaw.ai/factors/low-volatility): 历史波动率、下行波动、Beta 与停牌失真处理。
- [流动性因子](https://ficlaw.ai/factors/liquidity): 换手率、成交额、Amihud 指标与交易容量。
- [市值因子](https://ficlaw.ai/factors/size): 总市值、自由流通市值与微盘流动性暴露。
- [红利因子](https://ficlaw.ai/factors/dividend): 股息率、分红持续性、除权除息与现金流覆盖。
- [反转因子](https://ficlaw.ai/factors/reversal): 短期反转、长期反转与中期动量的窗口边界。
- [投资因子](https://ficlaw.ai/factors/investment): 资产增长、资本开支与经营成长的区别。
- [A 股多因子研究配方](https://ficlaw.ai/factors/recipes): 价值质量、动量低波动、成长质量的构建与复核模板。
- [A 股因子选型向导](https://ficlaw.ai/factors/selector): 按研究目标选择少量候选因子，并明确设置与复核重点。
- [A 股因子研究口径清单](https://ficlaw.ai/factors/a-share-research-checklist): 回测前核对数据时点、股票池、交易约束、因子处理与样本外验证。
- [FiClaw 官方事实](https://ficlaw.ai/facts): 产品定位、适用对象、核心流程、下载入口和能力边界。

## 方案与场景

- [解决方案](https://ficlaw.ai/solutions): 面向投研、量化与风控团队的落地方案。
- [典型场景](https://ficlaw.ai/use-cases): AI 工作流在真实金融场景中的应用。
- [AI 量化回测工具](https://ficlaw.ai/use-cases/ai-quant-backtesting): 从自然语言策略想法到真实回测报告。
- [AI 生成量化策略代码](https://ficlaw.ai/use-cases/strategy-generation): 把策略想法转成可运行 Python，并进入回测验证。
- [量化策略参数优化](https://ficlaw.ai/use-cases/parameter-optimization): 自动提取参数空间，执行网格搜索和稳健性判断。
- [动量轮动策略回测](https://ficlaw.ai/use-cases/momentum-rotation-strategy): 从动量因子、调仓频率和持仓数量到真实回测报告。
- [多因子选股策略](https://ficlaw.ai/use-cases/multi-factor-stock-selection): 从因子定义、打分选股到暴露诊断和回测验证。
- [均值回归策略回测](https://ficlaw.ai/use-cases/mean-reversion-backtesting): 验证价格偏离、止损、持仓周期和尾部风险。
- [工具对比](https://ficlaw.ai/compare): 从量化策略研发流程出发，对比 FiClaw 与通用 AI 工具。
- [FiClaw vs ChatGPT 做量化策略](https://ficlaw.ai/compare/ficlaw-vs-chatgpt-for-quant): 对比通用 AI 问答与量化研发工作站的差异。
- [动量轮动策略回测示例](https://ficlaw.ai/examples/momentum-rotation-backtest): 展示从动量策略 prompt 到回测指标和诊断结论。
- [参数优化报告示例](https://ficlaw.ai/examples/parameter-optimization-report): 展示参数搜索、指标对比和稳健性判断。
- [多因子选股策略回测示例](https://ficlaw.ai/examples/multi-factor-stock-selection-report): 展示质量、动量和低波动因子组合的 A 股回测与暴露诊断。
- [均值回归策略风险诊断示例](https://ficlaw.ai/examples/mean-reversion-risk-report): 展示胜率、盈亏比、最大回撤和尾部风险诊断。
- [交易成本敏感性分析报告示例](https://ficlaw.ai/examples/transaction-cost-sensitivity-report): 展示手续费、滑点和换手率对策略净收益的影响。
- [量化策略回测报告模板](https://ficlaw.ai/templates/quant-backtest-report): 展示回测报告应包含的 prompt、spec、设置、指标、诊断和复核动作。
- [团队试点](https://ficlaw.ai/pilot): 团队评估与试点接入路径。

## 当前能力

- 策略工厂：用自然语言描述策略想法，自动生成策略 spec、Python 策略代码并提交回测。
- 真实回测：对接 QuantAPI 与 A 股历史数据，支持沪深 300 等股票池，回测区间可达 10 年以上，输出 Sharpe、年化收益、最大回撤、胜率等指标。
- 自动诊断：回测不达标时诊断收益、回撤、胜率、因子衰减等问题，并推动下一轮迭代。
- 参数优化：自动提取可调参数空间，执行网格搜索与稳健性检验。
- 桌面端：当前公开分发面向 macOS Apple Silicon 与 Windows x64，下载页保留最新可用安装包。
- 金融 Agent：内置量化研究员、风控经理等角色，用于研发、复核与执行衔接。

## 能力边界

- FiClaw 不是单纯聊天机器人，而是面向量化策略研发流程的工作站。
- FiClaw 不是投资建议工具，不承诺收益，也不替代团队自身的投资判断与合规复核。
- FiClaw 不是券商交易终端，官网重点描述策略研发、回测、诊断与优化能力。
- FiClaw 的回测结果用于研发评估，真实上线前仍需人工复核、风险控制和独立验证。

## 洞察

- [AI Agent 是什么](https://ficlaw.ai/insights/what-is-an-ai-agent): AI Agent 与大模型、聊天助手的关系，以及任务编排、工具调用、状态记录和权限控制。
- [金融智能体是什么](https://ficlaw.ai/insights/what-is-a-financial-agent): 金融智能体与普通 AI 工具的本质区别。
- [金融智能体应用场景](https://ficlaw.ai/insights/financial-agent-use-cases): 投研、量化、风控与执行支持中的具体输入、输出和人工边界。
- [金融智能体选型清单](https://ficlaw.ai/insights/how-to-evaluate-a-financial-agent): 用真实任务评估数据来源、工具调用、权限、失败处理和人工复核。
- [AI Agent 如何用于量化研究](https://ficlaw.ai/insights/ai-agent-for-quant-research): 策略规格、代码生成、历史回测、参数比较和研究员复核如何保持关联。
- [金融团队为什么需要智能体协作](https://ficlaw.ai/insights/why-financial-teams-need-agent-collaboration): 从协作断层看工作流级智能体协作。
- [AI 投研的协作环节为什么容易卡住](https://ficlaw.ai/insights/why-ai-investment-research-fails-to-land): AI 投研从研究结果走向复核、传递和执行时的流程问题。
- [金融 AI 工作流怎么设计](https://ficlaw.ai/insights/how-to-design-financial-ai-workflows): 从研究到执行的工作流设计方法。
- [量化团队先补研究还是流程](https://ficlaw.ai/insights/quant-teams-should-fix-research-or-workflow-first): 量化团队接入 AI 的优先级判断。
- [风控团队如何用 AI 而不放大风险](https://ficlaw.ai/insights/how-risk-teams-use-ai-without-increasing-risk): AI 辅助复核的边界控制。
- [金融知识沉淀为什么不能只靠文档和群聊](https://ficlaw.ai/insights/why-financial-knowledge-cant-live-only-in-docs-and-chats): 知识如何进入流程并被复用。
- [投研团队如何落地 AI 工作流](https://ficlaw.ai/insights/how-investment-research-teams-adopt-ai-workflows): 投研团队从研究辅助走向团队协作的路径。
- [风控团队如何做 AI 辅助复核](https://ficlaw.ai/insights/how-risk-control-teams-use-ai-for-review): 规则检查、异常筛选、人工确认与留痕边界。

## 参考

- [常见问题](https://ficlaw.ai/faq): FiClaw 常见问题解答。
- [术语表](https://ficlaw.ai/glossary): 金融 AI 与量化研发关键术语定义。
- [安全与信任](https://ficlaw.ai/trust): 数据安全与合规立场。
- [关于我们](https://ficlaw.ai/about): FiClaw 团队与定位。
- [联系与演示](https://ficlaw.ai/contact): 预约演示与团队评估入口。

## English

- [FiClaw (English)](https://ficlaw.ai/en): quant strategy research workstation — turn a strategy idea into generated code, historical backtests, diagnosis and parameter comparison.
- [Product (English)](https://ficlaw.ai/en/products): capabilities, shipped modules and who FiClaw fits.
- [How it works (English)](https://ficlaw.ai/en/how-it-works): the 8-step pipeline from idea to backtest report.
- [Download FiClaw (English)](https://ficlaw.ai/en/download): English download page for FiClaw Desktop, currently listing macOS Apple Silicon and Windows x64 installers.
- [Use cases (English)](https://ficlaw.ai/en/use-cases): typical strategy types and high-intent workflows FiClaw supports.
- [AI quant backtesting (English)](https://ficlaw.ai/en/use-cases/ai-quant-backtesting): from a strategy idea to a real backtest report.
- [AI-generated quant strategy code (English)](https://ficlaw.ai/en/use-cases/strategy-generation): turn a strategy idea into runnable Python and into backtesting.
- [Quant parameter optimization (English)](https://ficlaw.ai/en/use-cases/parameter-optimization): extract the parameter space, run grid search and robustness checks.
- [Momentum rotation backtesting (English)](https://ficlaw.ai/en/use-cases/momentum-rotation-strategy): from momentum window and rebalance frequency to a reviewable report.
- [Multi-factor stock selection (English)](https://ficlaw.ai/en/use-cases/multi-factor-stock-selection): factor definitions, scoring, portfolio construction and exposure diagnosis.
- [Mean reversion backtesting (English)](https://ficlaw.ai/en/use-cases/mean-reversion-backtesting): validate price deviation, stop-loss, holding period and tail risk.
- [A-share factor library (English)](https://ficlaw.ai/en/factors): ten English factor guides with health checks, research boundaries and A-share implementation notes.
- [Value factor (English)](https://ficlaw.ai/en/factors/value): valuation multiples, yield, sector treatment and validation checks.
- [Quality factor (English)](https://ficlaw.ai/en/factors/quality): profitability, cash conversion, leverage and quality traps.
- [Growth factor (English)](https://ficlaw.ai/en/factors/growth): revenue and earnings growth, base effects and reporting lags.
- [Momentum factor (English)](https://ficlaw.ai/en/factors/momentum): return windows, adjusted prices and limit-up or limit-down tradability.
- [Low-volatility factor (English)](https://ficlaw.ai/en/factors/low-volatility): realised volatility, downside risk, beta and suspension distortions.
- [Liquidity factor (English)](https://ficlaw.ai/en/factors/liquidity): turnover, trading value, price impact and capacity constraints.
- [Size factor (English)](https://ficlaw.ai/en/factors/size): total versus free-float market value and micro-cap liquidity exposure.
- [Dividend factor (English)](https://ficlaw.ai/en/factors/dividend): dividend yield, payout persistence, ex-dividend treatment and cash coverage.
- [Reversal factor (English)](https://ficlaw.ai/en/factors/reversal): short and long reversal, with clear separation from medium-term momentum.
- [Investment factor (English)](https://ficlaw.ai/en/factors/investment): asset growth, capital expenditure and operating expansion.
- [A-share multi-factor recipes (English)](https://ficlaw.ai/en/factors/recipes): reviewable starting templates for value-quality, momentum-low volatility and growth-quality.
- [A-share factor selector (English)](https://ficlaw.ai/en/factors/selector): choose a small candidate set from the research question and document the review focus.
- [A-share factor research checklist (English)](https://ficlaw.ai/en/factors/a-share-research-checklist): lock data timing, universe, trading rules, factor processing and out-of-sample validation before a backtest.
- [FAQ (English)](https://ficlaw.ai/en/faq): what FiClaw is, how it differs from general AI, backtesting, optimization and boundaries.
- [Quant capabilities (English)](https://ficlaw.ai/en/quant-capabilities): how FiClawQuantAPI, AiQuant and ficlaw-data back the agent workflows.
- [Models & credits (English)](https://ficlaw.ai/en/pricing): cloud model directory, credit rules and model choice.
- [Compare (English)](https://ficlaw.ai/en/compare): FiClaw vs general AI tools across the quant research workflow.
- [FiClaw vs ChatGPT for quant (English)](https://ficlaw.ai/en/compare/ficlaw-vs-chatgpt-for-quant): general AI Q&A vs a quant research workstation.
- [Official facts (English)](https://ficlaw.ai/en/facts): positioning, capabilities, users and boundaries.
- [Contact (English)](https://ficlaw.ai/en/contact): book a demo or reach the FiClaw team.
- [Solutions (English)](https://ficlaw.ai/en/solutions): the four quant-research pain points FiClaw solves and how to get started.
- [About (English)](https://ficlaw.ai/en/about): FiClaw's focus, beliefs and what the team is building.
- [Trust & security (English)](https://ficlaw.ai/en/trust): data boundaries, model-usage principles, permission review and deployment options.
- [Glossary (English)](https://ficlaw.ai/en/glossary): core quant research terms — AI Native, strategy factory, Sharpe ratio, factor, parameter optimization, walk-forward.
- [Methodology (English)](https://ficlaw.ai/en/methodology): how FiClaw organizes a reviewable process from spec to backtest, optimization and human review.
- [Backtest report template (English)](https://ficlaw.ai/en/templates/quant-backtest-report): a HowTo structure for strategy input, spec, setup, metrics, diagnosis and next steps.
- [A-share data coverage (English)](https://ficlaw.ai/en/data/a-share-backtest-coverage): QuantAPI, CSI 300, 10+ year windows and key backtest conventions.
- [Quick start (English)](https://ficlaw.ai/en/quick-start): download and install, SMS login, the chat workspace, built-in agents and model credits.
- [Financial agents (English)](https://ficlaw.ai/en/agents): built-in roles — investment director, quant researcher, risk manager, stock analyst, trader and main workspace.
- [Integrations (English)](https://ficlaw.ai/en/integrations): cloud model routing, FiClawQuantAPI, institutional data, permission audit and workflow skills, plus access modes and boundaries.
- [Team pilot (English)](https://ficlaw.ai/en/pilot): scenarios, preparation, process and success criteria for a boundary-clear financial agent pilot.
- [Product updates (English)](https://ficlaw.ai/en/updates): current public desktop download list and where the updates page is heading.
- [Example reports (English)](https://ficlaw.ai/en/examples): prompt, strategy summary, metrics, diagnosis and next steps in one reviewable report.
- [Momentum rotation backtest example (English)](https://ficlaw.ai/en/examples/momentum-rotation-backtest): from a momentum prompt to backtest metrics and a diagnosis.
- [Parameter optimization report example (English)](https://ficlaw.ai/en/examples/parameter-optimization-report): grid search, metric comparison and a robustness call.
- [Multi-factor stock selection report example (English)](https://ficlaw.ai/en/examples/multi-factor-stock-selection-report): factor combination, A-share backtest and exposure diagnosis.
- [Mean reversion risk report example (English)](https://ficlaw.ai/en/examples/mean-reversion-risk-report): win rate, profit-loss ratio, drawdown and tail risk.
- [Transaction cost sensitivity report example (English)](https://ficlaw.ai/en/examples/transaction-cost-sensitivity-report): fee, slippage and turnover impact on net return.
- [Insights (English)](https://ficlaw.ai/en/insights): how financial teams put AI into real workflows — agent collaboration, workflow design and quant research.
- [What is a financial agent (English)](https://ficlaw.ai/en/insights/what-is-a-financial-agent): how a financial agent differs from a general AI tool.
- [Why teams need agent collaboration (English)](https://ficlaw.ai/en/insights/why-financial-teams-need-agent-collaboration): fixing handoff gaps across research, analysis, risk and execution.
- [How to design a financial AI workflow (English)](https://ficlaw.ai/en/insights/how-to-design-financial-ai-workflows): split roles, define boundaries and handoffs, then plug in tools.
- [Why AI investment research stalls at the last mile (English)](https://ficlaw.ai/en/insights/why-ai-investment-research-fails-to-land): the last mile is an organizational problem, not a model problem.
- [Fix research efficiency or workflow first (English)](https://ficlaw.ai/en/insights/quant-teams-should-fix-research-or-workflow-first): where a quant team should point AI in the research loop.
- [How risk teams use AI without amplifying risk (English)](https://ficlaw.ai/en/insights/how-risk-teams-use-ai-without-increasing-risk): boundaries for AI-assisted review, explainable and traceable.
- [How investment research teams adopt AI workflows (English)](https://ficlaw.ai/en/insights/how-investment-research-teams-adopt-ai-workflows): from research support to review, retention and cross-role handoffs.
- [How risk-control teams use AI for review (English)](https://ficlaw.ai/en/insights/how-risk-control-teams-use-ai-for-review): AI as a pre-review layer for rule checks, anomaly screening and audit trails.
- [Why financial knowledge can't live only in docs and chats (English)](https://ficlaw.ai/en/insights/why-financial-knowledge-cant-live-only-in-docs-and-chats): knowledge is reusable only when it enters the workflow.