trag-cli 2.1.3

Automatic Claude context builder for any coding project

T-RAG

Automatic Claude context builder for any coding project.

One command → scans your entire project → auto-detects what changed → builds a context summary → paste into Claude.ai → Claude instantly knows your full project state.

No Git needed. No weekly plan needed. No manual writing. Works for everyone.

Install

pip install trag-cli

That's it. No other setup needed.

Use

Go into any project folder and run:

trag

First time:

• Saves a baseline snapshot of all your files
• Builds full project context

Every time after:

• Automatically shows what changed since last run
• Builds updated context ready to paste into Claude

Options

trag                          # scan current folder
trag --project /my/project    # scan a specific folder
trag --plan todo.txt          # use a specific plan/todo file
trag --no-embed               # faster mode (skip ChromaDB)
trag --init                   # create optional notes file
trag --version                # show version

How change tracking works

T-RAG saves a snapshot of your file hashes inside trag_data/trag_snapshot.json in your project folder. Next run it compares and finds exactly what changed — no Git needed, fully local.

Each project has its own snapshot because trag_data/ is created inside each project folder.

Optional: weekly plan / todo

If you have a todo.txt, weekly_plan.txt, or tasks.txt in your project, T-RAG reads it automatically and includes your plan status in the context.

Optional: smarter AI reasoning (Ollama)

T-RAG works well without Ollama. For smarter AI-powered project summaries:

# Install Ollama from https://ollama.com then:
ollama pull qwen2.5-coder:7b

T-RAG auto-detects Ollama — nothing else to configure.