Json Mender Mcp

io.github.headlessherm-creator由 GitHub 账号 headlessherm-creator 发布v0.2.1更新于 Oct 30, 2026

Local MCP server: JSON repair, regex testing, JSON-Schema validation for AI agents.

已验证STDIO仅桌面开发者工具数据与分析

概览

AI 生成的概览

本地 MCP 服务器,可修复格式错误的 JSON、测试正则表达式,并按 JSON Schema 校验数据,全部为确定性的离线计算。

可以做什么
  • 修复损坏的 JSON(尾随逗号、未加引号的键、未转义的引号、被截断的结构)并返回解析结果
  • 用一个正则表达式匹配一个或多个字符串并返回匹配详情
  • 按 draft 2020-12 JSON Schema 校验 JSON 数据并给出逐字段的错误信息
适用场景
适用于助手或智能体反复生成无法解析的 JSON 时,需要确认生成的正则是否真的匹配目标字符串时,或在结构化输出被使用前需要按 schema 检查时。
运行要求
以 npm 本地进程方式运行,需要 Node.js v18 或更高版本。未声明需要 API 密钥、环境变量、请求头或网络访问。
安装前请注意
README 称不使用网络请求、遥测或凭据,工具仅在本地计算;但修复与校验结果仍应在采信为正确数据前人工复核。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Json Mender Mcp,将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。

其他 MCP 客户端

参照 仓库 中的启动说明。

README

JSON Mender MCP — Fix Malformed JSON, Validate JSON Schema, Test Regex for AI Agents

A free, local MCP server that fixes broken JSON output from LLMs, validates data against a real JSON Schema, and tests regex patterns deterministically — no API keys, no network calls, no telemetry.

If you're building an AI agent and keep hitting JSONDecodeError, unterminated strings, trailing commas, or unquoted keys in model output, this is the MCP server for that exact problem.

Same local-first, deterministic philosophy as polymath-megablaster-mcp, applied specifically to the "model wrote JSON that doesn't parse" problem.

The problem this solves

LLMs — especially smaller or local models (Llama, Qwen, Mistral) — routinely produce JSON that looks right but fails to parse: trailing commas, single quotes instead of double, unescaped inner quotes, missing commas, truncated output. Asking the model to "fix the JSON" usually just produces a second broken attempt. repair_json fixes it deterministically, every time, the same way, with no second model call needed.

Tools

  • repair_json — fix malformed/invalid JSON (trailing commas, unquoted keys, unescaped quotes, truncated structures) and return the corrected, parsed result. Built on jsonrepair.
  • validate_regex — test a regex pattern against one or more strings deterministically. Useful because models frequently write regex that compiles but doesn't match what they think it matches.
  • validate_against_schema — validate any JSON data against a real JSON Schema (via ajv, the most widely used JS Schema validator) and get precise, per-field, human-readable error messages instead of a generic pass/fail.

Setup

1. Install Node.js (v18+) if you don't have it: https://nodejs.org

2. Clone and install:

bash
git clone https://github.com/headlessherm-creator/json-mender-mcp.gitcd json-mender-mcpnpm install

3. Point your MCP client at it:

json
{  "mcpServers": {    "json-mender": {      "command": "node",      "args": ["/path/to/json-mender-mcp/src/index.js"]    }  }}

Setup docs for common clients:

Why local

No fetch/http import anywhere in src/tools.js — every tool runs as pure local computation using jsonrepair and ajv, both well-established, actively maintained npm packages. Nothing you send through this server leaves your machine.

Keywords

JSON repair MCP server, fix broken JSON from LLM, malformed JSON parser, JSON schema validation MCP, regex tester MCP, local MCP server no API key, Claude MCP JSON tools, AI agent structured output validation.

Related tools

Same local-first philosophy, different problems:

License

MIT

来源:README.md,提交 0f903b3

工具

3

repair_json

Repair malformed JSON (trailing commas, unescaped quotes, missing brackets) and return the parsed result.

输入参数text必填

validate_regex

Test a regex pattern against one or more strings deterministically and return match details.

输入参数flagspattern必填testStrings必填

validate_against_schema

Validate data against a real JSON-Schema (draft 2020-12) and report all errors.

输入参数data必填schema必填

版本历史

1
  1. v0.2.1最新Oct 22, 2026

评分

已满足适用检查项的 90%

该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。

  • 近 30 天内可连通25 分
  • 已声明认证方式10 分

    请在注册中心条目中声明凭据请求头,或保持端点能够响应探测,以便判定其认证方式。

  • 已列出工具必需15 分
  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分