Summary
When using GitHub Copilot Chat with Claude Code installed, it offers the ability to use Claude instead of GitHub Copilot as the backing Agent. Prompts entered in this mode error with Error: Claude CLI Error: Claude Code process exited with code 1. I realize this feature is in Preview, but thought you'd still want to know about this.
Possibly Related Context
I am running VSCode in a devcontainer with root.
I am not enabling dangerous permissions mode actively in any way. Claude Code does not run with dangerous permissions enabled when using it directly in the Claude Code extension.
Claude Code extension interface works just fine. This behavior is limited to the GitHub Copilot Chat interface.
Version Info
- Copilot Chat Extension Version: 0.44.1
- VS Code Version: 1.116.0
- OS Version: MacOS 15.7.1
- Feature (e.g. agent/edit/ask mode): Claude Code Agent
- Selected model (e.g. GPT 4.1, Claude 3.7 Sonnet): Claude Haiku 4.5
- Logs: Yes, below
Steps to Reproduce:
- Install extension
- Install Claude Code
- Install Claude Code Extension
- Authenticate with Anthropic using Claude Code Extension OAuth flow
- Open Copilot Chat window
- In bottom left, select "Claude" (instead of "Local")
- Send a prompt message
Developer Logs from GitHub Copilot Chat
2026-04-17 20:24:25.106 [info] Successfully activated the GitHub.vscode-pull-request-github extension.
2026-04-17 20:24:25.106 [info] [githubTitleAndDescriptionProvider] Initializing GitHub PR title and description provider provider.
2026-04-17 20:24:25.106 [info] Successfully registered GitHub PR title and description provider.
2026-04-17 20:24:25.106 [info] Successfully registered GitHub PR reviewer comments provider.
2026-04-17 20:24:26.766 [info] BYOK: Copilot Chat known models list fetched successfully.
2026-04-17 20:24:32.508 [info] [ClaudeLanguageModelServer] Claude Language Model Server started on http://localhost:42325
2026-04-17 20:24:32.991 [error] claude-agent-sdk stderr: --dangerously-skip-permissions cannot be used with root/sudo privileges for security reasons
2026-04-17 20:24:32.999 [error] [ClaudeRuntimeDataService] Failed to query agents from SDK: Error: Query closed before response received
2026-04-17 20:24:32.999 [error] Error: Claude Code process exited with code 1
at sGe.getProcessExitError (/root/.vscode-server/extensions/github.copilot-chat-0.44.1/dist/extension.js:328:21072)
at ChildProcess.r (/root/.vscode-server/extensions/github.copilot-chat-0.44.1/dist/extension.js:328:23968)
at Object.onceWrapper (node:events:634:26)
at ChildProcess.emit (node:events:531:35)
at ChildProcess.emit (node:domain:489:12)
at ChildProcess._handle.onexit (node:internal/child_process:293:12)
at Process.callbackTrampoline (node:internal/async_hooks:130:17)
Summary
When using GitHub Copilot Chat with Claude Code installed, it offers the ability to use Claude instead of GitHub Copilot as the backing Agent. Prompts entered in this mode error with
Error: Claude CLI Error: Claude Code process exited with code 1. I realize this feature is in Preview, but thought you'd still want to know about this.Possibly Related Context
I am running VSCode in a devcontainer with root.
I am not enabling dangerous permissions mode actively in any way. Claude Code does not run with dangerous permissions enabled when using it directly in the Claude Code extension.
Claude Code extension interface works just fine. This behavior is limited to the GitHub Copilot Chat interface.
Version Info
Steps to Reproduce:
Developer Logs from GitHub Copilot Chat