Preflight Checklist
What's Wrong?
When I start two Claude code sessions in the same folder, they seem to share history - which means they get mixed up with each others tasks. For example, if I start one Claude Code Session to do debugging and another to do feature development - the one will have the chat history of the other. After a particular task has been completed, the debugging instance starts to do Feature development because it seems to have access to the chat history of the other (e.g. if you use the up arrow when you are in the code interface you can up arrow to the prompts from the other instance). This means that Claude Code eventually stops following explicit instructions and does what it likes.
What Should Happen?
Each instance in the same project should be isolated from ther others. Even when resuming a chat --resume, the instance should be isolated and every chat interface should maintain its own context and chat history - the only time that things should be recalled is when you --resume an instance, then that instance's messaging, context, prompts etc should all be retrieved and used.
Error Messages/Logs
Steps to Reproduce
Open two instances of Claude code in the same project folder. Give each instance a task and watch as when one instance completes the task (often after a compaction) it will have context, prompting and history from the other instance.
Claude Model
Not sure / Multiple models
Is this a regression?
I don't know
Last Working Version
No response
Claude Code Version
1.0.115 (Claude Code)
Platform
Anthropic API
Operating System
Ubuntu/Debian Linux
Terminal/Shell
Other
Additional Information
No response
Preflight Checklist
What's Wrong?
When I start two Claude code sessions in the same folder, they seem to share history - which means they get mixed up with each others tasks. For example, if I start one Claude Code Session to do debugging and another to do feature development - the one will have the chat history of the other. After a particular task has been completed, the debugging instance starts to do Feature development because it seems to have access to the chat history of the other (e.g. if you use the up arrow when you are in the code interface you can up arrow to the prompts from the other instance). This means that Claude Code eventually stops following explicit instructions and does what it likes.
What Should Happen?
Each instance in the same project should be isolated from ther others. Even when resuming a chat --resume, the instance should be isolated and every chat interface should maintain its own context and chat history - the only time that things should be recalled is when you --resume an instance, then that instance's messaging, context, prompts etc should all be retrieved and used.
Error Messages/Logs
Steps to Reproduce
Open two instances of Claude code in the same project folder. Give each instance a task and watch as when one instance completes the task (often after a compaction) it will have context, prompting and history from the other instance.
Claude Model
Not sure / Multiple models
Is this a regression?
I don't know
Last Working Version
No response
Claude Code Version
1.0.115 (Claude Code)
Platform
Anthropic API
Operating System
Ubuntu/Debian Linux
Terminal/Shell
Other
Additional Information
No response