Skip to content
TraceXDocs
TraceX CLI

Coding mode

In a git repository the TraceX CLI works as a coding agent that reads, edits and runs your code, and also knows your production data, logs, errors and deploys.

In a git repository, tracex -c works like a coding agent: it lists, searches and reads the code, edits files and runs commands (tests, builds, git) to do what you ask. What makes it different from other coding agents is that it can also ask TraceX about production.

The problem it solves

Section titled “The problem it solves”

Coding agents only see the files on your machine. They can’t tell you whether a column is ever null in production, which code path customers actually hit, or what broke after yesterday’s deploy, so they guess. TraceX’s coding mode checks: it reads your local code and asks the TraceX project about the real data, logs, errors and deploys.

What it can do

Section titled “What it can do”
  • Answer questions about the code (summarize this project, where is signup handled?) from the files in your checkout.
  • Change the code: edit files and run commands such as tests, builds and git.
  • Ask about production: after tracex init, ask about the project’s databases, logs, errors, deploys and other repositories. Ask why did signups drop yesterday? and it asks TraceX; ask how does signup work? and it reads the local code. You see each question it sends to TraceX.

You approve every change

Section titled “You approve every change”
  • Every edit is shown as a diff, and every command before it runs. Answer y (yes), a (always, for this session) or n (no).
  • Without a terminal, for example in a script, edits and commands are refused.
  • Files outside the repository can’t be read or written, symlinks included.

AI usage

Section titled “AI usage”

The model runs through TraceX, so no AI keys are needed on your machine, and usage counts against your plan. TraceX doesn’t store the conversation.

Section titled “Related”