<< Go back to Posts

Github Copilot Agent VS Claude Code




Github Copilot VS Claude Code

Here are my first thoughts about Copilot VS Claude.

My experience is quite limited, I have played one month so far with Copilot, and one week (not full, I runned out of tokens ;) with Claude Code.

Pricing

First, there is a difference in Pricing which imply a difference is the way it is used / the autonomy the agent has to complete a task.

  • GC is per request: The agent limits is action to 25 iterations. At the end, you need to put a new coin in the machine
  • CC is per token: The agent has no incentive to stop quickly. You pay for what you consummed (ok, you have “global subscription”, but you have some limit of tokens per 5 hours).

This single difference makes the way we work with both different.

  • With Copilot, I am more likely to add details about the file it should read, the files it should edit, the test it shouldn’t write, so I efficiently use the 25 iterations.
  • Claude Code is more autonomous: I give it a task, maybe a few pointers, and it will run until completion.

UI

Next difference is User Interface:

GC is integrated into VSCode:

Pro:

  • we see the code edits, we can accept or reject, we see precisely the edit that has been done
  • we can run multiple agents simultaneously

Cons:

  • The debrief of the agent is in a very small font. Difficult to read, so I never take a look at that.

Claude Code runs directly in the terminal:

Pro

  • We are involved in the choices, giving our comments to the agent so it is guided in the correct direction
  • We need to review the code changes, so we know what are the edit made

Cons

  • When an edit has been made, it is difficult to rollback, plus it may impact other files

Default Behavior

With Both, you can customize Agent.md files so it follows your recommendation.

I haven’t enough experience with Claude Code, so I will focus on Github Copilot.

In the default mode, Copilot will write you tests and write / update Readme. I have developped a lot of streamlit GUI. “testing” a GUI cannot be done easily using test. So Agent was making tests for the different elements (not related to the GUI). This consummed me some time, but did not help me.

Documentation

Github Agent also like to write complex readme file. I like and dislike it at the same time. I dislike it because it is mainly to explain how the class developped work, how to use, …, but it seems irrelevant to me. It add “yet another file in the codebase”, while things are still “in developpement”, subject to change, so I may need to rewrite the readme. Plus, as I may need to edit the class, because some functionalities are useless, I would need to update the readme.

I like it because as the font size is too small in the chat window, getting a wrap-up readme let me time to read later (rather than retrieve the conversation I had with the agent).


>> You can subscribe to my mailing list here for a monthly update. <<