What file formats are supported?
JSON and newline-delimited JSON (NDJSON/JSONL). NDJSON files are automatically treated as an array of objects.
@zfirsty
基本信息
MCP server providing tools to query, inspect, and modify local JSON files.
使用下面的配置,将此服务器添加到你的 MCP 客户端。
{
"mcpServers": {
"jsonTools": {
"description": "Tools to query, inspect, and modify local JSON and NDJSON files.",
"command": "npx",
"args": [
"mcp-json-tools"
]
}
}
}Executes a JSONPath query on a local JSON or NDJSON file and returns an array of matching values. Reads both formats automatically; NDJSON is treated as an array of objects for querying. **Syntax Note**: Use `$.path...` for JSON objects and `$[selector]` (e.g., `$[*]` or `$[?(...)]`) for NDJSON arrays.
Executes a JSONPath query on a local JSON or NDJSON file and returns matching nodes (value + path). Reads both formats automatically; NDJSON is treated as an array of objects for querying. **Syntax Note**: Use `$.path...` for JSON objects and `$[selector]` (e.g., `$[*]` or `$[?(...)]`) for NDJSON arrays.
Reads a JSON or NDJSON file, executes JavaScript code within a sandboxed VM with file content ($1, object for JSON, array for NDJSON), lodash (_), and jsonpath (jp). Returns the result OR modifies the file (preserving original format: JSON or NDJSON) if the code's last expression is { type: 'updateFile', data: <new_data> }. **Do not use explicit `return` for the final result/update object; the last evaluated expression is used.** Has a 30s timeout. **WARNING: Executes user-provided code.**
Reads multiple JSON or NDJSON files, executes JS code within a sandboxed VM with file contents ($1 is array of parsed contents), lodash (_), and jsonpath (jp). Returns the result OR modifies files (preserving original formats) if the code's last expression is { type: 'updateMultipleFiles', updates: [...] }. **Do not use explicit `return` for the final result/update object; the last evaluated expression is used.** Has a 30s timeout. **WARNING: Executes user-provided code.**
MCP JSON Tools is an MCP server that lets you interact with local JSON and NDJSON (JSONL) files by querying with JSONPath and manipulating data via Lodash. It provides four tools—mcp_json_query, mcp_json_nodes, mcp_json_eval, and mcp_json_multi_eval—for developers using MCP‑compatible clients like Cursor or VS Code.
Configure your client to run the server via npx (recommended) or Node directly. In your client’s mcp.json (e.g., for Cursor or VS Code), set "command": "npx" and "args": ["mcp-json-tools"]. When using npx, all file paths passed to the tools must be absolute paths. Tools accept parameters such as file_path, json_path, js_code, and optionally count.
mcp_json_query)mcp_json_nodes)mcp_json_eval, mcp_json_multi_eval)vm module with configurable timeoutsJSON and newline-delimited JSON (NDJSON/JSONL). NDJSON files are automatically treated as an array of objects.
Use `mcp_json_eval` (single file) or `mcp_json_multi_eval` (multiple files). The last expression in your JavaScript code must be `{ type: 'updateFile', data: <new_data> }` (or `{ type: 'updateMultipleFiles', updates: [...] }`).
Lodash (as `_`) and jsonpath (as `jp`) are pre‑loaded in the sandboxed VM. The file content is available as `$1` (object for JSON, array for NDJSON).
Both `mcp_json_eval` and `mcp_json_multi_eval` have a 30‑second timeout.
No. The code runs inside Node.js’s `vm` module, which is safer than raw `eval()` but still poses risks of resource exhaustion or unintended logic. Only execute trusted code.
A one-of-a-kind resume builder that keeps your privacy in mind. Completely secure, customizable, portable, open-source and free forever. Try it out today!
A curated list of references for MLOps
YC (S26) | Open Computer History | Continuously record your company computer work, map your workflows, help you find work worth automating, and power your agents' context
This open-source curriculum introduces the fundamentals of Model Context Protocol (MCP) through real-world, cross-language examples in .NET, Java, TypeScript, JavaScript, Rust and Python. Designed for developers, it focuses on practical techniques for building modular, scalable,
Hindsight: Agent Memory That Learns
评论