Originally created by @MohamedBassem on GitHub (Apr 21, 2025).
Check Existing Issues
I have searched the existing issues and discussions.
Problem Description
MCP over MCPO was a great addition, however, I noticed that since I configured some tools, the LLM is constantly trying to use them even if it doesn't make sense. As an example:
When having the fetch MCP enabled, the model fabricates a stackoverflow URL, and try to fetch it to get an answer instead of just answering without using the tool.
Desired Solution you'd like
Similar to how we have toggles for web search and code interpreters, it would be great if by clicking on the tools icon, you can quickly toggle tools on and off per conversation.
Alternatives Considered
I've tried instructing the model (using a system prompt) to not use tools unless really needed, but haven't had great success.
Additional Context
No response
Originally created by @MohamedBassem on GitHub (Apr 21, 2025).
### Check Existing Issues
- [x] I have searched the existing issues and discussions.
### Problem Description
MCP over MCPO was a great addition, however, I noticed that since I configured some tools, the LLM is constantly trying to use them even if it doesn't make sense. As an example:

When having the fetch MCP enabled, the model fabricates a stackoverflow URL, and try to fetch it to get an answer instead of just answering without using the tool.
### Desired Solution you'd like
Similar to how we have toggles for web search and code interpreters, it would be great if by clicking on the tools icon, you can quickly toggle tools on and off per conversation.

### Alternatives Considered
I've tried instructing the model (using a system prompt) to not use tools unless really needed, but haven't had great success.
### Additional Context
_No response_
@MohamedBassem commented on GitHub (Apr 21, 2025):
@tjbck Not sure which setting are you referring to. User settings? That I understand but it's global. I was thinking about a per conversation setting similar to the Web search and Code interpreter ones.
@MohamedBassem commented on GitHub (Apr 21, 2025):
@tjbck Not sure which setting are you referring to. User settings? That I understand but it's global. I was thinking about a per conversation setting similar to the `Web search` and `Code interpreter` ones.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @MohamedBassem on GitHub (Apr 21, 2025).
Check Existing Issues
Problem Description
MCP over MCPO was a great addition, however, I noticed that since I configured some tools, the LLM is constantly trying to use them even if it doesn't make sense. As an example:
When having the fetch MCP enabled, the model fabricates a stackoverflow URL, and try to fetch it to get an answer instead of just answering without using the tool.
Desired Solution you'd like
Similar to how we have toggles for web search and code interpreters, it would be great if by clicking on the tools icon, you can quickly toggle tools on and off per conversation.
Alternatives Considered
I've tried instructing the model (using a system prompt) to not use tools unless really needed, but haven't had great success.
Additional Context
No response
@tjbck commented on GitHub (Apr 21, 2025):
You can toggle them off from the settings.
@MohamedBassem commented on GitHub (Apr 21, 2025):
@tjbck Not sure which setting are you referring to. User settings? That I understand but it's global. I was thinking about a per conversation setting similar to the
Web searchandCode interpreterones.