Mqttable product overview: a full MQTT workbench

Mqttable is a local-first MQTT workbench for engineers who need to operate, inspect, reproduce, and explain MQTT behavior. It starts with the basics: connect clients, publish, subscribe, inspect packets, analyze captures, replay scenarios, run benchmarks, monitor runtime events, and use local utilities without sending artifacts to random web tools. Its strongest differentiator is MCP, but MCP is not the whole product. The product value is the whole loop: live operation, packet evidence, deterministic replay, fault injection, load testing, runtime diagnosis, desktop runtime, and an agent surface that can read and verify the same evidence.

Main surfaces

• Connections for live MQTT clients, subscriptions, publishes, trace rows, resend, and scheduled messages.
• Trace and PCAP Analysis for packet evidence, findings, diagnostic checks, and replay candidates.
• Replay for turning a trace into a named scenario with dry run, checks, run trace, and reports.
• Proxy for controlled network fault injection and recovery checks.
• Bench for load tests with preflight, live charts, history, comparison, and reports.
• Monitor for runtime metrics, events, issue groups, and operational context.
• Tools for local hash, gzip, JWT, QR, diff, image, and time utilities.
• Desktop and Remote Runtime for local-first work with an explicit hosted runtime path.
• MCP and CLI for agent-assisted discovery, planning, confirmation, and verification.

How to read the product

Do not treat Mqttable as a chatbot shell around MQTT. Treat it as an MQTT workbench that also exposes the right structured surface to agents. A human can stay in the UI. An agent can use MCP. Both should point back to the same broker state, trace evidence, replay run, benchmark result, or monitor event.