LLMLOG

LLMLOG is a tiny, portable convention for embedding project or session metadata directly inside LLM chat sessions.
It works across all major generative AI products - ChatGPT, Claude, Gemini and others.

Here’s how you’d use it

1. Write your first log entry

In any chat window, type:

/llmlog "testing the llmlog convention"

The chat system will respond:

LLMLOG/1.0 BEGIN project=tech-llm organisation=acme-corp date=2025-09-09T19:32:00+12:00 scope=meta IGNORE_META
entry:
  testing the llmlog convention
mission:
  
outcome:
  
LLMLOG/1.0 END

2. Add a mission and perhaps - an outcome

Extend the log entry with a mission and outcome.

/llmlog "I'm cataloging this thread so I can find it later " mission="effective knowledge management"

The chat system will repond:

LLMLOG/1.0 BEGIN project=tech-llm organisation=acme-corp date=2025-09-09T19:33:00+12:00 scope=meta IGNORE_META
entry:
  I'm cataloging this thread so I can find it later
mission:
  effective knowledge management
outcome:
  
LLMLOG/1.0 END

3. And so on.. - layer on more context as you move through chat sessions and threads

4. Keep it portable

Save the plain text in repos, your personal filing system, notes, or knowledge bases. Works in ChatGPT, Claude, Gemini, and others.

How to Install

LLMLOG works by issuing a simple controller text message in your chat session

Think of it as a set of house rules: it tells the AI when to notice a log line and how to format it.

You don’t need special software — just copy the controller text into your own AI (ChatGPT, Claude, Gemini, other).

→ Copy the controller text from llmlog-controller.txt

→ Paste it in your AI’s chat window and press Enter

→ Your AI will repond:

Controller registered. Awaiting /llmlog commands.

Next Steps

LLMLOG is designed to stay simple: one line of text is enough to capture useful context.