Enterprise Research Evidence Chain
LocalAbout Enterprise Research Evidence Chain
A read-only MCP server for enterprise investment and industry research, serving frameworks, frozen snapshots and historical factor summaries. Each result includes sources, an as-of date and falsification conditions; it does not provide trading advice.
Key features
- 01Research framework excerpts
- 02Frozen stock snapshots
- 03Qualified frozen hypotheses
- 04Historical theme radar summaries
- 05Historical factor IC summaries
- 06Safe single-stock analysis
Use cases
- 01Reviewing sourced company research
- 02Checking historical factor summaries
MCP support
Exposes research-reading and controlled analysis tools to MCP clients. The wrapper validates inputs, forwards tenant credentials to a research data service, and filters trading-related fields from outputs.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
uvx diyan-research-mcpDIYAN_API_KEYSecretTenant credential issued by Diyan. Fill this on the MCP detail page.DIYAN_TRANSPORTstdio, sse, or streamable-http. Defaults to sse in hosted environments; use stdio for local clients.Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEdiyan-research-mcp · 企业研究证据链
面向企业投研与产业研究团队的只读研究 MCP,提供研究框架、冻结快照与历史因子摘要,每条结论附来源、数据截至日与证伪条件,不输出交易建议。
这个仓库里没有数据
本仓库只包含转发壳,职责是三件事:校验入参、携带租户凭据调用研究数据服务、在返回前剥掉交易类字段。研究框架、冻结快照、假设台账和因子摘要都存放在服务端,不随代码分发。
AI 客户端 ──MCP──▶ diyan-research-mcp(本仓库) ──HTTPS──▶ 研究数据服务
入参校验 / 凭据转发 / 输出过滤 数据、租户、档位、配额、审计
这样分层有三个后果,都是有意为之:
- 壳可以公开审阅,因为它不含任何专有数据,也不含任何密钥。
- 租户身份、档位和配额一律由服务端裁定,壳不做信任判断,也就无法被绕过。
- 输出过滤留在壳内而不是只放在服务端,即便服务端出错或被篡改,交易类表述也无法离开进程。
工具
| 工具 | 说明 |
|---|---|
framework_excerpt | 读取研究框架片段,可选 enterprise / business / finance / people / brand |
stock_snapshot_latest | 读取最新冻结的股票研究快照 |
hypothesis_get | 读取通过交付证据门槛的冻结假设 |
cognition_radar_read | 读取历史主题研究雷达摘要 |
ic_backtest_summary | 读取冻结的因子历史 IC 摘要 |
research_analyze_safe | 调用受控研究接口做单标的研判 |
工具是否可用取决于凭据对应的档位。免费档开放研究框架与部分冻结快照,其余返回 error_type: backend 并说明需要付费租户凭据。
配置
唯一必填项是 DIYAN_API_KEY,在 MCP 详情页填写即可。其余变量见 .env.example,只在自建部署或联调时需要。
凭据缺失时进程仍会正常启动,但每次工具调用都会返回明确的配置错误。这是刻意的:托管环境可能在使用者填写凭据之前就拉起进程,启动即崩溃会让详情页上的服务看起来是坏的。
本地运行
python -m pip install -r requirements.txt
export PYTHONPATH=src
export DIYAN_API_KEY=<你的租户凭据>
export DIYAN_TRANSPORT=stdio
python -m diyan_research_mcp.server
测试不需要网络和凭据:
python -m unittest discover -s tests -v
边界
- 只读。没有任何写入、下单、部署或数据导出能力。
- 不返回实时行情。快照与摘要都是冻结产物,每条结果都带
as_of。 - 不生成买卖、仓位、股数、跟单或收益承诺。含此类表述的提问在发出请求前就会被拒绝。
- 不读取使用者的本地文件、本地数据库或非必要环境变量。
服务端契约
如需自建后端,接口定义见 docs/BACKEND_CONTRACT.md。
许可与归属
由深圳瞳桦文化传媒有限公司开发与维护。MIT 许可,见 LICENSE。
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.