Kiro Product
IDE
Feature Description
Feature Description
When multiple MCP servers are configured in Kiro, the total number of available tools grows rapidly. This triggers warnings like "too many tools" and can degrade agent performance and context efficiency.
Proposed solution: Allow users to organize MCP servers into named groups, and selectively activate a group per chat session.
Problem
- Adding several MCP servers quickly inflates the tool count (e.g., 5 servers × 10 tools = 50+ tools in context).
- Kiro already warns about excessive tool counts, which signals this is a real constraint.
- There is currently no way to scope which MCP tools are active for a given task or conversation.
Proposed Behavior
- Group creation – Users can create named groups (e.g., "File Operations", "Database", "Web") and assign one or more MCP servers to each group.
- Group activation in chat – When starting or during a chat, the user can specify which group(s) to use (e.g.,
#FileOperations), so only the tools from those servers are loaded into context.
- Ungrouped servers – MCP servers that are not assigned to any group continue to work exactly as they do today (always active), preserving full backward compatibility.
- Multiple groups – A user should be able to activate more than one group at a time if needed.
Example
Groups:
📁 File Operations → [filesystem-mcp, git-mcp]
🌐 Web → [browser-mcp, search-mcp]
🗄️ Database → [postgres-mcp]
Ungrouped (always active): [memory-mcp]
In chat: "Using the File Operations group, refactor all TypeScript files in /src"
Benefits
- Reduces tool count in context → fewer warnings, better agent focus.
- Gives users explicit control over what capabilities are available per task.
- Scales gracefully as the MCP ecosystem grows and users add more servers.
Additional Context
This is especially relevant for power users who rely on many MCP servers for different workflows (e.g., frontend work vs. backend work vs. DevOps tasks). Grouping would let them switch "tool profiles" without manually enabling/disabling servers each time.
Use Case
I work on multiple types of projects — frontend, backend, and DevOps. For each context I use different MCP servers. Right now, all of them are active at once, which floods the tool list and triggers "too many tools" warnings.
With grouping, I'd create a "File & Git" group for daily coding tasks and a "Database" group for data work. When I start a chat, I'd pick the relevant group so only those tools are in context — keeping the agent focused and avoiding the tool limit warnings.
Additional Context
No response
Kiro Product
IDE
Feature Description
Feature Description
When multiple MCP servers are configured in Kiro, the total number of available tools grows rapidly. This triggers warnings like "too many tools" and can degrade agent performance and context efficiency.
Proposed solution: Allow users to organize MCP servers into named groups, and selectively activate a group per chat session.
Problem
Proposed Behavior
#FileOperations), so only the tools from those servers are loaded into context.Example
Groups:
📁 File Operations → [filesystem-mcp, git-mcp]
🌐 Web → [browser-mcp, search-mcp]
🗄️ Database → [postgres-mcp]
Ungrouped (always active): [memory-mcp]
In chat: "Using the File Operations group, refactor all TypeScript files in /src"
Benefits
Additional Context
This is especially relevant for power users who rely on many MCP servers for different workflows (e.g., frontend work vs. backend work vs. DevOps tasks). Grouping would let them switch "tool profiles" without manually enabling/disabling servers each time.
Use Case
I work on multiple types of projects — frontend, backend, and DevOps. For each context I use different MCP servers. Right now, all of them are active at once, which floods the tool list and triggers "too many tools" warnings.
With grouping, I'd create a "File & Git" group for daily coding tasks and a "Database" group for data work. When I start a chat, I'd pick the relevant group so only those tools are in context — keeping the agent focused and avoiding the tool limit warnings.
Additional Context
No response