Kiro IDE Tracing with Neural Inverse

What is Kiro? Kiro is an AI-powered IDE by AWS that features AI agents to help you write, edit, debug, and understand code. Kiro provides a hooks system that lets you run custom scripts at different lifecycle points during AI agent activity.

What is Neural Inverse? Neural Inverse is an open-source LLM engineering platform. It helps teams trace LLM applications, debug issues, evaluate quality, and monitor costs in production.

What Can This Integration Trace?

Using Kiro Hooks, this integration automatically captures AI agent activity and sends traces to Neural Inverse. All 10 Kiro hook types are supported:

Kiro Hook Type	What It Traces
Prompt Submit	User prompts and queries
Agent Stop	Agent completion with status scores
Pre Tool Use	Tool invocations before execution
Post Tool Use	Tool results and duration
File Create	New file creation
File Save	File modifications
File Delete	File deletions
Pre Task Execution	Spec task start
Post Task Execution	Spec task completion
Manual Trigger	On-demand trace events

Key observability features include:

Traces grouped by conversation: Each conversation gets its own trace in Neural Inverse
Sessions grouped by workspace: All traces from the same workspace folder are grouped into a session
Dynamic tagging: Automatic tags based on tool type, file operations, and model
Completion status scoring: Scores (0-1) recorded on agent completion
Non-blocking error handling: Tracing never interferes with your coding flow

How It Works

Kiro triggers hook event
    → Shell command runs hook-handler.js
    → Reads input from stdin + environment variables
    → Creates/updates Neural Inverse trace for the conversation
    → Routes to event-specific handler
    → Creates spans, generations, scores in Neural Inverse
    → Flushes data before exit

Trace Hierarchy

Each conversation produces a trace with the following structure:

Trace — one per conversation
- Session — grouped by workspace folder
- Generations — user prompts
- Spans — tool use, file operations, task execution
- Events — agent stop, manual triggers
- Scores — completion status (0-1)

Quick Start

Set Up Neural Inverse

Sign up for Neural Inverse Cloud or self-host Neural Inverse.
Create a new project and copy your API keys from the project settings.

Clone the Example Repository

Clone the langfuse-examples repository:

git clone https://github.com/langfuse/langfuse-examples.git

Copy Hook Files to Your Project

Copy both the Kiro hook definitions and the handler code to your project:

cp -r langfuse-examples/applications/kiro-langfuse/.kiro/hooks/ your-project/.kiro/hooks/
cp -r langfuse-examples/applications/kiro-langfuse/hooks/ your-project/hooks/

This adds:

.kiro/hooks/ — Kiro hook definition files (one per hook type)
hooks/ — The Node.js handler code and Neural Inverse client

Install Dependencies

cd your-project/hooks && npm install

Configure Neural Inverse Credentials

Create a .env file in your project root with your Neural Inverse API keys:

# Neural Inverse credentials - get these from https://cloud.langfuse.com
LANGFUSE_SECRET_KEY=sk-lf-...
LANGFUSE_PUBLIC_KEY=pk-lf-...

# Optional: self-hosted Neural Inverse URL (defaults to cloud.langfuse.com)
# LANGFUSE_BASE_URL=https://your-langfuse-instance.com

Environment Variables:

Variable	Description	Required
`LANGFUSE_PUBLIC_KEY`	Your Neural Inverse public key	Yes
`LANGFUSE_SECRET_KEY`	Your Neural Inverse secret key	Yes
`LANGFUSE_BASE_URL`	Neural Inverse base URL. EU: `https://cloud.langfuse.com`, US: `https://us.cloud.langfuse.com`, Japan: `https://jp.cloud.langfuse.com`, HIPAA: `https://hipaa.cloud.langfuse.com`	No (defaults to EU)

Enable Hooks in Kiro

Open the Hook UI in Kiro:

Mac: Cmd + Shift + P → "Kiro: Open Kiro Hook UI"
Windows/Linux: Ctrl + Shift + P → "Kiro: Open Kiro Hook UI"

The hook files in .kiro/hooks/ will appear automatically. Toggle on the hooks you want to trace.

Each hook file follows this format:

{
  "version": "1.0.0",
  "enabled": true,
  "name": "Neural Inverse - Prompt Submit",
  "when": {
    "type": "promptSubmit"
  },
  "then": {
    "type": "runCommand",
    "command": "KIRO_HOOK_EVENT=prompt_submit node hooks/hook-handler.js"
  }
}

View Traces in Neural Inverse

Open your Neural Inverse dashboard and navigate to Traces. You can:

Filter by tag kiro to see all Kiro traces
Filter by session name (your workspace folder) to see traces from a specific project
Click a trace to see the full conversation breakdown with nested spans, generations, and scores

Customization

Filter Specific Tools

Edit .kiro/hooks/langfuse-pre-tool-use.kiro.hook and update the "toolName" field to target specific tools. The following built-in categories are supported:

read — all built-in file read tools
write — all built-in file write tools
shell — all built-in shell command-related tools
web — all built-in web tools
spec — all built-in spec tools
* — all tools (built-in and MCP)

You can also use prefix filters to target tools by source:

@mcp — all MCP tools
@powers — all Powers tools
@builtin — all built-in tools

Prefixes starting with @ are matched by regex, so you can use patterns like @mcp.*sql.* to match specific MCP tools by name.

You can also configure tool hooks via the Kiro Hook UI — type each tool name and press Enter to add it.

Filter Specific Files

Edit .kiro/hooks/langfuse-file-save.kiro.hook and change "patterns": ["**/*"] to target specific files:

["src/**/*.ts"] — TypeScript source files
["**/*.js", "**/*.ts"] — JS and TS files
["**/*", "!node_modules/**"] — exclude node_modules

Prerequisites

Node.js v18+
Kiro IDE installed
A Neural Inverse account (cloud or self-hosted)

Resources

Was this page helpful?

On this page