Material Cleaning MCP
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Material Cleaning MCP
独立的材料清洗工具,使用标准 MCP stdio,也可通过命令行运行。功能是清理采集头、网页标记、人工确认的导航行和无关段落,并保护引文、重算重复出现次序、输出清洗审计。作者 Shalom,版本0.1.0。
本工具独立于 NVivo,可用于论文材料、网页文本和其他定性研究输入。原件由调用方保留,输出为新文本和审计。默认保留源码、运行记录、JSON/XML等技术证据;CSV按完整记录转为字段阅读视图。
安装和接入
git clone https://github.com/youngiry/material-cleaning-mcp.git
cd material-cleaning-mcp
.\install.ps1安装器输出 config.local.json,将 mcpServers.material_cleaning 合并到支持 MCP 的客户端。也可使用 Python 3.11+ 创建虚拟环境并安装 requirements.txt,然后以该解释器启动 server.py。代码采用标准库清洗核心,Windows已实测,其他操作系统可先运行测试。
DeepSeek、Qwen、Kimi等模型通过支持本地 MCP 的 Harness/客户端使用该工具。工具本身无需模型 API key;具体客户端配置格式以其版本为准。
Related MCP server: genpark-document-redaction-pii-scrubber-skill
输入和输出
调用 materials_clean(documents, passages, rules)。documents 为 name/text 对象列表;passages 为 source/quote/occurrence/label 对象列表;occurrence从0开始,表示原文中第几次出现。rules按来源名称指定 remove_lines 或 keep_lines,行号从1开始且指向本次输入原文。
先阅读各来源并确认正文边界,再给出规则。自动规则仅处理常见采集头与格式;正文取舍仍依材料内容判断。被引文覆盖的删除行会受到保护并列入审计。输出 documents、passages 及 audit,审计包含前后哈希和字符数。完整示例见 examples/input.json。
.\.venv\Scripts\python.exe cleaner.py --input examples/input.json --output results/cleaned.json
.\.venv\Scripts\python.exe verify_live.pyCLI要求新的输出文件路径,以保留原件。规则与原件版本成对保存;重新清洗时从原件开始。当前范围为已经提取的文本;PDF版面识别、扫描件OCR和网页下载由上游处理。
与 NVivo 组合
先用本工具清洗,再把 documents映射为NVivo的sources、passages映射为codings中的quote和occurrence。NVivo建工程、编码及保存由独立的NVivo Headless MCP完成。组合步骤见 docs/workflow.md。
DeepSeek Harness专用示例见 接入说明 和 examples/dsh.patch.yml。
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Free deterministic PDF, Markdown, citation, invisible-character, and formatting cleanup for agents.
Analyse and clean up text: stats, language, keywords, summary, sentiment, PII redaction, diff.
Inspect and remove PDF/DOCX author metadata and comments; approve tracked changes before cleaning.
Dedupe, flatten and clean messy JSON rows (emails, phones, URLs, HTML) in one call, as JSON or CSV.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceEnables cleaning tabular data by detecting duplicates, placeholder nulls, type issues, spelling variants, encryption damage, outliers, and personal data, then applying safe fixes automatically while requiring human approval for risky changes.MIT
- FlicenseNot gradedqualityNot gradedmaintenanceEnables document layout parsing, complex table reconstruction, formula auditing, and PII/credential redaction with Luhn verification, reversible salt hashing, and audit logs through MCP-compatible clients.-
- AlicenseNot gradedqualityBmaintenanceEnables autonomous web browsing and extraction by canonicalizing URLs, detecting anti-crawler traps such as circular paths and infinite pagination, pruning DOM content, mapping forms, converting tables to JSON, and evaluating agent trajectories.7MIT
- AlicenseNot gradedqualityBmaintenanceEnables cleaning clinical-research tables by mapping fields, classifying and normalizing cells, routing confidence-gated decisions to a review queue, and exporting reports and audit logs through an MCP stdio interface.MIT