Triage by Chodeberg

Product Introduction

1. Definition: Triage by Chodeberg is a developer operations (DevOps) and software quality assurance (QA) automation platform. It is a bug reporting and triage system that integrates with AI coding assistants via the Model Context Protocol (MCP) to create self-reproducing, executable bug reports.
2. Core Value Proposition: It exists to eliminate the "works on my machine" problem in software development by automating the creation of deterministic, sandboxed bug reproductions. Its primary value is in filtering out non-reproducible issues, saving developer time on triage and ensuring that only verified, actionable bug reports reach the engineering team.

Main Features

1. AI-Assisted Bug Filing via MCP: Customers report bugs directly from their AI assistant (e.g., Claude Code) using a provided MCP server. The agent automatically detects the app's environment and version, then opens a live session in a matching, isolated sandbox to drive the app and capture the failure. This feature automates the initial, often messy, bug reporting step.
2. Deterministic Replay on Clean Machines: The core technical engine. Every submitted bug report is automatically replayed multiple times on pristine, ephemeral virtual machines that match the exact build (e.g., v1.18.35). A report is only forwarded to developers if it consistently reproduces the crash or failure across all clean replays. This ensures report validity.
3. Comprehensive Telemetry & Flight Recorder: Each replay session is instrumented as a "flight recorder." It captures granular telemetry data sampled every second, including container memory (RSS) and peak usage, CPU utilization per process, full process trees (highlighting orphans), exit codes, signals, core dumps, and a full screen recording with keystroke/resize steps. This transforms subjective descriptions into quantitative evidence.

Problems Solved

1. Pain Point: The inefficiency and ambiguity of traditional bug reports. Reliance on user-written steps, screenshots, and videos leads to time-consuming back-and-forth, unreproducible issues, and a lack of concrete system state data (memory, CPU) at the moment of failure.
2. Target Audience: Development teams building and shipping Command-Line Interfaces (CLIs), Terminal User Interfaces (TUIs), Linux desktop applications, and software libraries/SDKs. It is particularly valuable for open-source maintainers and teams with a user base that utilizes AI coding assistants.
3. Use Cases: Essential for: 1) Receiving high-fidelity bug reports from external users or community members. 2) Automating the first layer of QA triage for intermittent or environment-specific crashes. 3) Providing engineering and support teams with immediate, replayable proof of a bug, complete with performance profiling data to diagnose memory leaks or CPU spikes.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):
   • Pros: Drastically reduces triage time by guaranteeing reproducibility. Provides deep, automated telemetry that is otherwise manual to collect. Integrates seamlessly into modern AI-assisted developer workflows via MCP. Security-focused design with offline sandboxes and no secret exposure.
   • Cons: Primarily targets CLI/TUI/Linux app ecosystems, less immediately relevant for pure web or mobile apps. Requires integration (connecting repo, setting up builds) for full functionality. Relies on users reporting via supported AI agents, potentially missing reports from other channels unless a public form is used.
2. Key Alternatives & Differentiation:
   • Traditional Issue Trackers (GitHub Issues, Jira): Triage differentiates by automating the evidence collection and verification before the ticket is created. Instead of a text description, the ticket includes a live, replayable environment and telemetry.
   • Screen Recording Tools (Loom, Veed): These capture what happened visually. Triage captures the full system context (memory, CPU, processes) alongside the visual, and automatically replays it in the correct environment for verification.
   • Automated Testing Suites: While testing suites run pre-defined tests, Triage is for reactive, user-reported issues. It creates a deterministic test case from a user's live session, bridging the gap between production incidents and QA.

Frequently Asked Questions (FAQ)

1. How does Triage by Chodeberg handle security and prevent malicious code execution? Reports execute in ephemeral, offline sandboxes with no internet access. Builds only reach allowlisted hosts, and no user secrets enter the environment. The system runs the reporter's steps against your own verified build, not arbitrary code.
2. What is the pricing model for Triage's bug reproduction service? Triage uses a credit-based system where 1 credit ≈ 1¢. Consumption is metered per minute of sandbox runtime (builds, live sessions, replays). It offers 100 free credits monthly and sells prepaid credit packs ($100, $250, $500, $1000) that never expire.
3. Can Triage integrate with my existing CI/CD pipeline and issue tracking workflow? Yes. It connects directly to your GitHub repository to build environments from your tags. Verified reports generate a detailed README with steps and data, which can be linked to or integrated into external trackers. Its MCP server also allows coding agents to pull reports directly.
4. What types of applications is Triage best suited for? Triage is specifically optimized for teams shipping Command-Line Interfaces (CLIs), Terminal User Interfaces (TUIs), Linux desktop applications, and software libraries. Its sandbox environment is tailored to these contexts, though the core concept applies to any software with reproducible builds.
5. How does the "clean machine" replay work to verify a bug? When a bug is reported, the Triage system automatically launches multiple fresh, identical virtual machines based on the exact application build version cited. It replays the recorded steps in each isolated environment. A bug is only considered verified and forwarded to you if it consistently fails in the same way across all clean replays, eliminating machine-specific state issues.