bug - github co-pilot keeps telling me it is in read only mode when it is in agent mode

[roblatour]

🏷️ Discussion Type

Bug

💬 Feature/Topic Area

Code Search and Navigation

Body

every time i write a prompt in agent mode the copilot ai does a bunch of reading and then tells me it is in read-only mode and can't make any changes.

I then need to tell the ai:
"you are in not in read only mode, you are in agent mode"

after which time it sometimes says it agrees and continues with the change in write mode, other times I have to tell it to write a temp file and then delete it before it will continue on forward

Here is a screenshot

[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. ⭐

[luacys]

Copilot Agent mode says it's "read-only" but edits work once you push it: what's happening and how to fix it

The screenshot is the key clue: after being told to write and delete a temp file, the agent did create and delete it. So the edit tools were available the whole time. The model wrongly concluded it was read-only. That narrows things down to a few causes, all fixable or at least diagnosable.

Several people are hitting this, across models (GPT-5.x, Claude, Gemini):

• Agent mode selected but Copilot behaves as read-only Ask mode after VS Code update microsoft/vscode#312212: Agent mode selected, Copilot claims read-only Ask mode (still open)
• Copilot Agents have lost Tool access despite Tools enabled microsoft/vscode#335438: edit tool missing from the tool list sent to the model
• Agent mode exposes only read-only tools in WSL; agent-host URI errors microsoft/vscode#342756: VS Code 1.142 / Copilot Chat 0.70, read-only tools in Agent mode (opened this week)

Likely causes

1. Edit tools aren't in the model's first tool list. Newer VS Code builds load some tools on demand (virtual tools / tool_search), especially when you have many MCP servers or extensions. The model reads files, doesn't see an edit tool, and decides the session is read-only. When you insist, it searches for or calls the tool and it works. This matches the behavior in the screenshot exactly.
2. Leftover context from Ask or Plan mode. If the chat started in Ask or Plan mode, or forked from one, earlier turns tell the model it's read-only. Switching the dropdown doesn't erase that history.
3. A custom agent or instructions restricting tools. A .agent.md file (in .github/agents/ or your user profile) with tools: [read, search], or instructions saying "don't modify files", can leak into Agent sessions. #342756 suspects exactly this.
4. Files are actually read-only to VS Code. Check whether files.readonlyInclude or files.readonlyFromPermissions is set, or whether the files are locked on disk.

Step 1: See what the model actually received (2 minutes)

Run the failing prompt, then open the Command Palette and run Show Chat Debug View. Expand the request and look at the tools list.

• No edit tool (apply_patch, replace_string_in_file, create_file or similar): the tools were never sent. This is a VS Code / Copilot bug or a config problem; continue with Step 2.
• Edit tool present but the model still claims read-only: the model is misreading its context. Skip to Step 3.

Step 2: Fix the tool list

• Click Configure Tools (the tools icon in the chat input). Make sure the built-in edit tools are enabled for the session, not just in global settings.
• Confirm the picker shows the built-in Agent, not a custom agent with a similar name. Check .github/agents/ and your user-level agents for anything that restricts tools.
• If you have many MCP servers, temporarily disable the ones you don't need for this task. This keeps the tool count low enough that edit tools aren't deferred.
• Update VS Code and the Copilot Chat extension together; Copilot Chat only supports the latest VS Code.

Step 3: Stop the model second-guessing itself

Start a fresh chat in Agent mode rather than switching modes mid-conversation.

Then add a standing instruction to .github/copilot-instructions.md. It does the same job as the temp-file trick, automatically:

## Tool access
You are running in Agent mode with file-editing and terminal tools.
Never state that the session is read-only or that you cannot edit files
unless an edit tool call has actually failed. If you are unsure, call
tool_search for edit tools, or attempt the edit, before concluding you
lack access. When you decide on a change, apply it directly.

If it still happens on one model, try another for the same prompt. If only one model does it, mention that in your bug report.

Step 4: Report it with a log (this is what gets it fixed)

The VS Code team has asked for request logs on these issues, not screenshots. Follow the guide at https://github.com/microsoft/vscode/wiki/Copilot-Issues#language-model-requests-and-responses and add the log to microsoft/vscode#312212, or open a new issue linking it. Include:

• VS Code version, Copilot Chat version, OS (and whether you're on WSL, SSH or a container)
• The model selected
• Whether the edit tool appeared in the Chat Debug tool list
• Number of MCP servers enabled and any custom agents

That last detail separates "tools not sent" from "model ignored the tools", which is exactly what the maintainers need to triage it.

[sr-vijay]

Step 1: See what the model actually received (2 minutes)
Step 2: Fix the tool list
Step 3: Stop the model second-guessing itself