MCP JSON Tools

Overview

MCP JSON Tools is a lightweight MCP server that empowers AI assistants to interact seamlessly with local JSON and NDJSON (newline‑delimited JSON) files. By combining the expressive power of Lodash for data manipulation and JSONPath for querying, it turns ordinary flat files into dynamic, queryable resources that can be read, inspected, and transformed on demand. This eliminates the need for external databases or complex ETL pipelines when an assistant needs to pull structured data from logs, configuration files, or exported reports.

The server exposes a set of high‑level tools that cover the full data lifecycle:

• Querying – lets you retrieve specific values using standard JSONPath syntax. It automatically normalises NDJSON files into an array of objects, so the same query language works across both formats.
• Inspection – returns not only the matched values but also their exact paths within the structure, enabling context‑aware processing or debugging.
• Evaluation & Modification – and run arbitrary JavaScript in a sandboxed Node.js VM, with Lodash () and JSONPath () pre‑loaded. This allows complex transformations or analytics to be performed directly on the file, while preserving its original format (JSON or NDJSON).
• Safety – All eval tools are bounded by configurable timeouts and run in a restricted VM, mitigating the risk of runaway scripts or resource exhaustion.

These capabilities are especially valuable for developers building AI‑driven workflows that require real‑time access to structured data stored locally. Typical use cases include:

• Data exploration – Quickly pull out specific metrics or logs for analysis without writing custom parsers.
• Configuration management – Let an assistant read and adjust JSON‑based configuration files, applying changes in a controlled manner.
• Report generation – Combine multiple NDJSON streams (e.g., event logs) into aggregated summaries or dashboards.
• Testing and debugging – Inspect the internal state of JSON files produced by other tools or services, helping to diagnose issues in complex pipelines.

Integration with AI assistants is straightforward: the server registers its tools under the MCP namespace, and a client can invoke any tool by name, passing file paths and query expressions as arguments. The results are returned in JSON format, ready for the assistant to incorporate into its responses or further processing steps.

By unifying querying, inspection, and transformation in a single, safe, and easy‑to‑deploy MCP server, JSON Tools removes the friction of working with local structured data, enabling developers to focus on higher‑level logic while letting AI assistants handle the nitty‑gritty data manipulation.