Skip to content

About

A debugging tool built for agent workflows

DebugBundle is designed around one idea: agents and humans both debug faster when the input artifact is structured, compact, and deterministic.

Modern production systems generate enormous volumes of logs, traces, and error reports. The challenge is not collecting signals — it is turning those signals into actionable debugging context.

DebugBundle captures production incidents, normalizes and groups events by fingerprint, assembles deterministic debug bundles, and delivers them through API, CLI, and MCP. The same bundle that a human developer reads in the terminal is the same artifact an AI agent processes programmatically.

The project is built with a local-first architecture. You can capture, process, and inspect incidents entirely on your machine with zero cloud dependency. The hosted platform adds team collaboration, longer retention, remote probes, and automated improvement analysis — but the core debugging workflow is always available locally.

Interface parity

API, CLI, and MCP are first-class equals. No capability is dashboard-only. Automation is not a second-class integration.

Deterministic output

Given the same events, DebugBundle produces byte-identical bundles. No random IDs, no wall-clock timestamps. Reproducible debugging artifacts every time.

SDK safety

SDKs never throw uncaught exceptions, block your request cycle, or crash the host process. Failures are caught internally and swallowed silently.

Learn more