Questions about Kintsugi
What is Kintsugi?
Kintsugi is a local-first safety layer and command firewall for AI coding agents like Claude Code, Cursor, Codex, Gemini, and Copilot. It intercepts dangerous shell commands before execution, explains them in plain language, and allows one-command undo. The software maintains a tamper-evident audit log and operates without cloud dependency or telemetry.
How does the command parsing engine work?
The software parses real shell structure using a fast tokenizer and a true bash AST parser written in pure Rust. It fails toward caution by holding any line the parser cannot fully understand, ensuring zero catastrophic-classified-as-safe commands across a golden corpus of 176 dangerous commands.
What are the two tiers of the decision engine?
Kintsugi uses a deterministic rule engine that decides Allow, Hold, or Deny without ever consulting a model. A second tier can optionally use local CPU GGUF inference to provide a risk score, but the gate decision remains rule-based and impossible to prompt-inject.
Which specific command categories does Kintsugi block?
The catastrophic categories map to data destruction like rm -rf and DROP TABLE, disk writes like dd and mkfs, and secret reads including .env and ~/.ssh keychains. The system verified 0 crashes across 1.4M fuzz inputs while maintaining a zero-cloud architecture.
How do I install Kintsugi?
The tool is distributed as an installable application via a shell script from its website. Users can copy the installation command 'curl -fsSL https://kintsugi.tools/install.sh | sh' to get started immediately.
Who is the ideal user for this tool?
Kintsugi is designed for AI agents that run real shell commands on a computer, such as those used in development workflows. It specifically targets scenarios where agents might execute irreversible deletes or destructive database operations.
What happens when a dangerous command is caught?
The system holds commands that are ambiguous, like deleting a local build folder, or catastrophic, like removing an entire source directory. It allows users to approve once or always allow these actions after reviewing the plain-language explanation.
How does the audit log ensure integrity?
The tamper-evident timeline records every event with a hash chain, so tampering with a past entry breaks verification. This ensures the record serves as evidence rather than just a convenience for the user.
Why is the block decision deterministic?
The block decision is deterministic and cannot be talked out of by a clever prompt because it relies on fixed rules a human wrote. A model can only explain and add caution but never downgrades a verdict or opens the gate.
How does Kintsugi handle complex shell structures?
The software includes a golden corpus that guarantees zero catastrophic-classified-as-safe commands. It catches commands hidden inside substitutions like echo "$(rm -rf /)" or here-strings like bash <<<'rm -rf /'.
What are the core architectural guarantees?
The system operates entirely offline with no cloud dependency or telemetry. Every screen is a faithful render of the real desktop app, and the model runs locally.
What testing and licensing information is available?
The MITRE ATT&CK + GTFOBins corpus was used to verify the safety rules. The tool is MIT licensed and has been operating since at least 2026.
How fast is the interception process?
The website states that the system runs in well under a millisecond for safe commands. It only interrupts for the ones that can actually hurt you, keeping calm until it must shout.
What is the official website of Kintsugi?
The official website of Kintsugi is kintsugi.tools.
Does Kintsugi have reviews on Trustburn?
Kintsugi has no reviews on Trustburn yet. Anyone who has used it can write the first review on its Trustburn page.
Post your question publicly — other visitors, and whoever manages this page, can answer. It will be visible on this page.
Want to hear about this company again? Follow it and we’ll email you when a new review appears.
Based on reviews, listing data and the company's website. Verify anything important with the company directly.
Questions & Answers
Which AI coding agents like Cursor or Copilot are officially supported for this firewall?
Trustburn Team suggests a starter question on new profiles — the business hasn’t answered yet.
Sign in to post
Your text is saved — you'll come right back to post.
Don't have an account? Create one
News & Press
No news yet. This is where the business shares its own updates and press.
Photos
No photos yet. Photos shared by reviewers will show up here.
About Kintsugi
Kintsugi is a local-first safety layer and command firewall designed for AI coding agents such as Claude Code, Cursor, Codex, Gemini, and Copilot. It intercepts dangerous shell commands before they execute, explains them in plain language, and makes them reversible through one-command undo functionality. The software maintains a tamper-evident audit log of agent activity and operates without cloud dependency or telemetry. Kintsugi is distributed as an installable application, with installation performed via a shell script from its website.
- Website
- kintsugi.tools
