IDEA intellij - copilot cannot see the output in terminal of command it executed

[dantuch]

Select Topic Area

Bug

Copilot Feature Area

Copilot Agent Mode

Body

I made copilot try a basic:
echo "This is a test output to see if I can view it"
and it didnt see the output in my terminal - is that normal?

"I see the issue now. It appears that while I'm executing commands in the terminal, I'm not able to see their actual output content. This applies not just to Spectral but to basic commands like echo as well. The function_results are showing me that commands were executed, but the actual output text isn't being displayed within the Output: section.
This explains why I've been having trouble verifying the Spectral validation results - I can confirm that commands are running but I can't see their actual textual output in the terminal.
Given this limitation, let's modify our approach. Instead of relying on direct terminal output visibility, we can use the read_file function to check the contents of the validation output "

copilot: 1.5.45.10445-243-nightly
IDEA: Build #IU-251.25410.129, built on May 9, 2025
tested on Mac: 15.0.1 (24A348)

btw. my copilot in MOST cases doesnt see the terminal output - but sometimes it does see it - so it seems not deterministic to me.

[Akash1134]

Hi @dantuch,

Thank you for reporting this issue with GitHub Copilot in agent mode not being able to reliably see terminal output in IntelliJ IDEA. This is a known limitation in the current implementation of Copilot's agent mode for JetBrains IDEs.

Issue Details

From your report, you're experiencing:

• Copilot agent mode cannot consistently see the output of terminal commands it runs
• You tested with a basic command (echo "This is a test output to see if I can view it") with the same result
• The behavior appears to be intermittent - sometimes it can see the output, but most times it cannot

Technical Explanation

This behavior occurs due to how Copilot agent mode interfaces with the JetBrains terminal. When Copilot executes a command in the terminal:

1. It can confirm the command was executed (which explains the "function_results" confirmation you mentioned)
2. However, the mechanism for capturing the actual terminal output text in JetBrains IDEs is not consistently reliable

The intermittent nature of the issue (sometimes working, sometimes not) is likely related to timing and buffering differences depending on the specific terminal session, command type, and output volume.

Current Workarounds

Until this is addressed in a future update, here are some effective workarounds:

1. File-based Output Redirection:

   • Modify commands to redirect output to files that Copilot can then read
   • Example: echo "Test output" > output.txt && cat output.txt

2. Alternative Command Verification Methods:

   • For validation operations (like your mentioned Spectral validation), have Copilot check for exit codes instead of trying to read output text
   • Use commands that create artifacts Copilot can verify independently

3. Manual Output Checking:

   • For critical terminal operations, you may need to check the output yourself and provide the results back to Copilot in your next prompt

Thank you for your patience and for helping us improve GitHub Copilot!

[rnorth]

@Akash1134 thanks for these workarounds, but they're pretty intrusive and aren't practical to do for every single agent operation.

I notice it's now a few months since this discussion was raised and it's still a problem. Please could you provide some clearer updates on when this will be tackled?

[Dave-Vaughan]

Is still the case after updating this morning. vscode can do this just fine - infact sometimes its scary how close to a professional hacker the speed of it is...., but i really dont want to use vscode as i'm used to the tools in intellij.

[github-actions[bot]]

💬 Your Product Feedback Has Been Submitted 🎉

Thank you for taking the time to share your insights with us! Your feedback is invaluable as we build a better GitHub experience for all our users.

Here's what you can expect moving forward ⏩

• Your input will be carefully reviewed and cataloged by members of our product teams. 
  • Due to the high volume of submissions, we may not always be able to provide individual responses.
  • Rest assured, your feedback will help chart our course for product improvements.
• Other users may engage with your post, sharing their own perspectives or experiences. 
• GitHub staff may reach out for further clarification or insight. 
  • We may 'Answer' your discussion if there is a current solution, workaround, or roadmap/changelog post related to the feedback.

Where to look to see what's shipping 👀

• Read the Changelog for real-time updates on the latest GitHub features, enhancements, and calls for feedback.
• Explore our Product Roadmap, which details upcoming major releases and initiatives.

What you can do in the meantime 💻

• Upvote and comment on other user feedback Discussions that resonate with you.
• Add more information at any point! Useful details include: use cases, relevant labels, desired outcomes, and any accompanying screenshots.

As a member of the GitHub community, your participation is essential. While we can't promise that every suggestion will be implemented, we want to emphasize that your feedback is instrumental in guiding our decisions and priorities.

Thank you once again for your contribution to making GitHub even better! We're grateful for your ongoing support and collaboration in shaping the future of our platform. ⭐

[amoerie]

I've had this issue today on Windows. Changing the terminal engine to "Classic" and the shell to "CMD.exe" fixed things for me.

[rpalacios]

I have same problem too.

I tried it with engine "Classic" and powershell 7, and it worked fine.

[Arham4]

In VS Code (in the context of Java) and running tests I generally tell copilot to "use test runner for java to run tests." Then it can read the results of tests. Wonder what the alternative is for IntelliJ.

[hmrguez]

Does anyone have a workaround for macos? Not very keen on using powershell for the time being

[davidcarma]

Solution: Fix Copilot Agent's Terminal Output Issue on macOS
Problem:
When using GitHub Copilot Agent in Visual Studio Code on macOS, the agent may fail to read terminal command output, causing it to hang or provide incorrect responses. This is often related to the default shell being zsh.

Solution:
The issue can be resolved by changing the default terminal shell in VS Code from zsh to bash. This provides an environment that the Copilot Agent can reliably interact with.

Step-by-Step Instructions
Open the Command Palette in VS Code by pressing Cmd+Shift+P on your keyboard.

In the search bar, type Terminal: Select Default Profile.

From the list of available profiles, select "bash".

Close and reopen any active integrated terminals to apply the change.

After following these steps, the next time you use the Copilot Agent, it will use the bash shell, which should allow it to correctly read command output.

Why This Works:
While both bash and zsh are powerful shells, they have subtle differences in how they handle command output, scripting, and I/O redirection. The Copilot Agent's underlying logic may have a more robust or specific integration with bash that is not fully compatible with zsh, leading to the observed behavior. Switching to bash provides a known working environment for the agent.

[HWiese1980]

It seems in IntelliJ IDEA I can trigger this bug by actively using the Copilot terminal session for manual commands myself. I can "fix" the issue by closing the session and let Agent open a "clean" new one.

[brolland-byg]

Having the problem with bash under wsl-linux-Ubuntu-bash
it is getting annoying because starting a new session costs a new premium request and loose the history.
Is there any way to step back to an older plugin version ?

[yuv-c-netapp]

Sorry @Akash1134 and Intellij - but the suggestions (pressing enter after a cmd, or changing theme to zsh) are either degrading the capabilities in agent mode or restrict the dev from using its own environment.
For example, podman isn't accessible in my bash shell, but is from zsh.
So this needs to be addressed.
Please see the answers in this issue (vs code). I am personally not using p10k, so I don't think its relevant for me, but its a starting point.

Please fix this. This is a paid product and should support working with the most common shells and coding assitants.

[lmhajdukd]

Same issue, Copilot doesn't see npm test results - failing tests and stacktraces after asking it to run them. I'm not using p10k, pure bash. Still doesn't work. Paying for a license for no reason.

[ChrisDugdale]

Same for the the command runs, then just hangs waiting until I cancel it. This is in Rider using copilot agent mode.