Do not stop agent reasoning if tool fails #148
Description
Another issue I encountered is that the bot, when in agent mode, stops any reasoning efforts when a tool call fails. This is inconvenient, since for some MCP servers (especially for the Azure DevOps MCP server that I am using), the bot may confuse itself about the arguments for a tool call and run into a tooling error. In these cases (for my use case) it would be preferable to allow the bot to reason about the encountered error in order to fix the arguments, rather than hard stopping the whole reasoning process.
Example:
