nats-view 0.2.1

NATS JetStream topic watcher

nats-view

A command line utility for watching NATS and NATS JetStream topics with support for filtering, pretty-printing JSON, and historical replay.

Features

• Real-time Monitoring: Watch core NATS subjects with ease.
• JetStream Support: Replay historical data using --since, --deliver-all, or --deliver-last.
• Visualization: Pretty-printed JSON, colorized output, timestamps, and headers.
• Filtering: Server-side subject filtering and client-side regex payload filtering.
• Security: Support for User/Pass, Token, Credentials, and TLS/SSL.
• MCP Compatible: Ready to be used as a tool for LLMs via the Model Context Protocol.

Installation

Via pipx (Recommended)

pipx install .

Via Poetry (Development)

poetry install
poetry run nats-view --help

Usage

Basic

Watch all subjects on localhost:

nats-view

Watch specific subjects:

nats-view "orders.>" "logs.*"

JetStream History

Replay messages from the last hour:

nats-view --js --since 1h "orders.>"

Get the last known value for a subject:

nats-view --js --deliver-last "status.system"

Filtering & formatting

Show only messages containing "error", pretty-print the JSON, and exit after 5 matches:

nats-view --filter "error" --pretty --count 5 ">"

MCP Tool Usage

This tool includes a definition for the Model Context Protocol (MCP), allowing AI assistants to inspect your NATS clusters.

Tool Definition

The tool definition is located in mcp-tool.json.

Configuration Example (Claude Desktop / Generic MCP Client)

To use this with an MCP server that runs CLI commands (like a generic "command-runner" MCP server), map the arguments as follows:

Executable: nats-view (ensure it is in your PATH)

Argument Mapping:

• subjects -> Positional arguments
• host -> -h
• port -> -p
• count -> -n
• filter -> --filter
• js -> --js
• since -> --since
• deliver_all -> --deliver-all
• deliver_last -> --deliver-last
• timestamp -> --timestamp
• headers -> --headers

Note for LLMs: When using this tool, always specify a count (e.g., 10 or 50) to prevent the tool from running indefinitely when watching high-traffic subjects.