Outputs

tCode's output during a turn is not a single block of text at the end; it is a running transcript of every step as it happens, followed by a short summary once the turn is done. This page shows what that actually looks like.

Tool call headers

Each tool call gets a one-line header as soon as it starts, naming the tool and the one detail that distinguishes it from any other call to the same tool, a file path, a search pattern, the command itself:

Permission prompts inline

Where a tool needs your approval first, that appears inline before it runs, showing exactly what is about to happen:

The closing response

Underneath that final message, a short activity line recaps the turn in one glance, which files were read, which were edited, and roughly how much changed, so scrolling back through the full transcript is not the only way to see what happened.

What a read-only step actually prints

A read-only step, a file read or a search, prints its header the same way an edit does, but nothing more beneath it by default: the result feeds straight back into the model's next decision rather than being dumped onto the screen in full, since a raw file's contents or a long list of search matches is rarely something you need to read yourself in the moment, only something the model needs.

Shell command exit codes, the one exception

A shell command is the one exception to that: its exit code is always shown directly under the command header, colored to make success or failure obvious at a glance, since that is the one piece of information about a command's result that is genuinely useful to see immediately, independent of whatever the model goes on to say about it afterward.

When a tool call fails

If a tool call fails, whether a file cannot be found, a command exits non-zero, or something else goes wrong, that failure is shown plainly as part of the transcript rather than hidden, and the model's next step responds directly to it, trying something else, asking you a question, or explaining what went wrong.

When a turn declines to do something

When a turn declines to do something, because you answered no to a permission prompt, or because plan mode recorded an action instead of running it, that also appears directly in the transcript rather than being folded invisibly into the final response, so it is always clear afterward exactly which of the things tCode wanted to do actually happened and which did not.

Syntax highlighting in responses

Code shown inside a response, whether quoted from an existing file or newly written, is syntax-highlighted using whichever theme is currently active (see Configuration), the same highlighting used for the code blocks throughout this documentation, so a snippet reads the same way it would in your editor rather than as plain, undifferentiated text.