I hope this discussion finds you well! I wanted to bring up an important issue that many of us developers are encountering: the current limitation on MCP (Model Context Protocol) tools that’s affecting our development workflows.
Current Situation
Many users are currently seeing an “Exceeding total tools limit” warning when working with MCP tools in Cursor. This limitation is significantly impacting productivity, especially for developers who rely on multiple MCP tools for comprehensive development workflows.
Why This Matters
MCP tools have revolutionized how we interact with various services and systems directly from our development environment. They enable us to:
Enhanced Productivity: Access to multiple tools simultaneously allows for seamless workflow integration
Better Context Management: Different tools serve different purposes - file management, API interactions, database queries, etc.
Comprehensive Development: Complex projects often require multiple specialized tools working in harmony
The Problem
The current tool limit prevents developers from:
Using comprehensive tool suites that require 10+ tools
Leveraging the full potential of MCP ecosystem
Building complex, multi-service applications efficiently
Maintaining productive development workflows
Proposed Solutions
Increase the Default Limit: Raise the current limit to accommodate modern development needs (perhaps 25-50 tools)
Tiered Limits: Implement different limits based on subscription tiers:
Free users: Current limit
Pro users: Higher limit (25-30 tools)
Enterprise users: Even higher or unlimited
Dynamic Loading: Allow selective tool loading/unloading during development sessions
Configuration Options: Let users configure which tools are most critical for their workflow
Impact on Community
This enhancement would:
Improve developer satisfaction and productivity
Better leverage the growing MCP ecosystem
Support more complex development workflows
Attract more developers to the Cursor platform
Community Input Needed
I’d love to hear from other developers:
How is the current limit affecting your workflow?
What would be an ideal number of MCP tools for your use case?
Are there specific tool combinations that are essential for your work?
Conclusion
Increasing the MCP tools limit would be a significant quality-of-life improvement for the developer community. It would enable us to fully utilize the powerful MCP ecosystem that Cursor has helped create.
Thank you for considering this request, and I look forward to hearing thoughts from both the team and community!
somewhat related to this request, is there a way to have different mcp tools for different projects/repos?
Especially for supabase where I have to put in my project name and my access token in the mcp.json I have to manually change that everytime I switch projects.
My wish/request would be to have the mcp.json in the repo under .cursor/mcp or something like that. Is there a way to achieve this currently or are there any plans to have specific MCP connections for specific projects?
When Cursor forwards 40+ raw MCP tools, the LLM wastes context judging tools, so latency increases and precision drops. Disabling tools manually is tedious and error-prone.
Alternatively: The paper “RAG-MCP: Mitigating Prompt Bloat in LLM Tool Selection via Retrieval-Augmented Generation” (https://arxiv.org/abs/2505.03275) embeds every tool’s metadata into a vector index and retrieves only the K most relevant entries before prompting. Authors report >50 % token cuts and ~3× higher tool-selection accuracy.
Two Practical Paths
Raise the cap – simplest to ship but keeps prompts fat and costly.
Retrieval proxy – place an external proxy that does RAG-based pre-selection and exposes just high-level endpoints. This flow is already implemented in the open-sourcemcpproxy-go (https://github.com/smart-mcp-proxy/mcpproxy-go/) if you need relief today.
Yeah this seems limiting. You should not have to worry about adding more tools. Am i supposed to toggle on and off? I only get it to call tools when i say so?
In the first implementation of MCPs you could toggle MCPs on/off but not tools. MCP servers provide information on the MCP capabilities and this has to be included in each AI API request.
This leads to:
Higher consumption of tokens as each API request has to include higher amount of information
Errors in AI responses as increased context with task unrelated information confuses AI.
Potential security issue from exposed access to sensitive information or destructive commands.
Therefore toggles for MCP tools have been added to allow you to avoid both cases.