MCP Configuration
MCP tools and permission prompts
Tools exposed by an MCP server go through the same permission system as any other tool — the first use of a new one typically prompts for approval.
Connecting a server doesn't automatically grant blanket trust for everything it can do; each distinct tool it exposes is still subject to the same approve-or-deny flow as built-in tools, at least on first use.
This is worth expecting rather than being surprised by — a newly connected server whose tools keep prompting isn't malfunctioning, it's behaving the same way a new shell command would the first time you run it.
See /permissions for adjusting how a specific MCP tool is treated going forward, the same way you'd scope any other repeatedly-used command.
Approve the prompt once, then adjust via /permissions if it should be trusted going forwardFrequently asked questions
What does MCP tools and permission prompts do in Claude Code?
Tools exposed by an MCP server go through the same permission system as any other tool — the first use of a new one typically prompts for approval.
How do I use it?
Try: Approve the prompt once, then adjust via /permissions if it should be trusted going forward
Any caveats to know about?
See /permissions for adjusting how a specific MCP tool is treated going forward, the same way you'd scope any other repeatedly-used command.
Is this covered in the Claude Code Toolkit?
Yes — MCP tools and permission prompts and patterns like it are part of the full command, hook, and template pack included with a Claude Code Toolkit subscription.
Want the full command, hook, and template pack?
This is one pattern out of the full Claude Code Toolkit — subscribe for the complete set, plus everything we add going forward.
Subscribe to the toolkit — $79/mo