# As a condition of accessing this website, you agree to abide by the following
# content signals:

# (a)  If a Content-Signal = yes, you may collect content for the corresponding
#      use.
# (b)  If a Content-Signal = no, you may not collect content for the
#      corresponding use.
# (c)  If the website operator does not include a Content-Signal for a
#      corresponding use, the website operator neither grants nor restricts
#      permission via Content-Signal with respect to the corresponding use.

# The content signals and their meanings are:

# search:   building a search index and providing search results (e.g., returning
#           hyperlinks and short excerpts from your website's contents). Search does not
#           include providing AI-generated search summaries.
# ai-input: inputting content into one or more AI models (e.g., retrieval
#           augmented generation, grounding, or other real-time taking of content for
#           generative AI search answers).
# ai-train: training or fine-tuning AI models.
# use:      how AI systems may consume the content (immediate, reference, or full).

# ANY RESTRICTIONS EXPRESSED VIA CONTENT SIGNALS ARE EXPRESS RESERVATIONS OF
# RIGHTS UNDER ARTICLE 4 OF THE EUROPEAN UNION DIRECTIVE 2019/790 ON COPYRIGHT
# AND RELATED RIGHTS IN THE DIGITAL SINGLE MARKET.

# BEGIN Cloudflare Managed content

User-agent: *
Content-Signal: search=yes,ai-train=no,use=reference
Allow: /

User-agent: Amazonbot
Disallow: /

User-agent: Applebot-Extended
Disallow: /

User-agent: Bytespider
Disallow: /

User-agent: CCBot
Disallow: /

User-agent: ClaudeBot
Disallow: /

User-agent: CloudflareBrowserRenderingCrawler
Disallow: /

User-agent: Google-Extended
Disallow: /

User-agent: GPTBot
Disallow: /

User-agent: meta-externalagent
Disallow: /

# END Cloudflare Managed Content

          
 
 
   
 
Visit the Blog →
 
  
 
 
NEW
 
📝 Latest Article:
 
Deep Dive: How MCPProxy Uses MCP Tool Annotations for Smarter Routing
 
 
 

MCPProxy.app is the Open Source Smart Proxy for AI Agents
 

Works with any MCP-compatible client: Cursor IDE, Claude Desktop, ChatGPT, and more.
 

A desktop application that extends AI-agent productivity by connecting hundreds of MCP servers 
          with intelligent tool discovery, token optimization, and advanced security quarantine protection.

✨ Works offline • 🖥️ User-friendly desktop app • 🔓 100% Open Source
 
  
 
  Download Latest Installers 
  
  
 
🪟 Windows AMD64
 
Setup Installer for Windows
 
  Signed by SignPath Foundation 
 
  
  
  
 
🍎 macOS Apple Silicon
 
DMG Installer for M1/M2/M3/M4 Macs
 
✨ Signed & Notarized • Recommended
 
  
  
  
 
🐧 Linux — apt / dnf
 
Signed package repositories for Debian / Ubuntu / Fedora / RHEL
 

sudo apt install mcpproxy   ·   sudo dnf install mcpproxy
 
✨ amd64 + arm64 • auto-updates via apt/dnf
 
  
 
 
View all installation methods (Homebrew, AUR, ARM64, tarballs) →
 
 
  
 
  Quick Start 
  
 
 
1
 
Install DMG
 
 

Download and install the DMG package above. Launch MCPProxy from Launchpad or Applications folder—it will run quietly in your system tray.
 
  
 
 
2
 
Add to Cursor IDE
 
 
  
Install into Cursor
 
 

One-click install directly into Cursor IDE via deeplink
 
 
  
  
Documentation
  
View on GitHub
  
Download Binary
 
 
 
 
  
 
 
  
 
  
 
Scale Beyond API Limits
 

Connect hundreds of MCP servers while bypassing Cursor IDE's 40-tool limit and
OpenAI's 128-function API cap. Provider limits
no longer constrain your AI productivity.
 
  
 
  
 
Speed Up & Save Tokens
 

Intelligent tool discovery eliminates function schema bloat. RAG-MCP research
shows ~99% token savings while maintaining 43% accuracy improvements over baseline methods.
 
  
 
  
 
Advanced Security Protection
 

Automatic quarantine system protects against Tool Poisoning Attacks (TPAs) and malicious MCP servers. 
            All new servers are quarantined for security review before activation.
 
 
 
 
   
 
 

Native System Integration
 
  
 
 
System Tray - Upstream Servers
 
    
Click to view full size
 
 
 
 
System Tray - Quarantine Management
 
    
Click to view full size
 
 
 
  
 
 
Cursor IDE Integration
 
    
Click to view full size
 
 
 
 
Tool Security Analysis
 
    
Click to view full size
 
 
 
 
 
 
  
 
 
 

How MCPProxy Works
 

MCPProxy acts as an intelligent federating gateway, consolidating multiple MCP servers 
          behind a single smart endpoint with built-in security and optimization.
 
 
  
 
1
 
Server Discovery
 

MCPProxy automatically discovers and indexes tools from all configured upstream MCP servers, 
            building a comprehensive searchable catalog.
 
  
 
2
 
Intelligent Search
 

When AI agents need tools, they query MCPProxy with natural language. 
            BM25 search returns the most relevant tools from the entire catalog.
 
  
 
3
 
Security Review
 

New servers are automatically quarantined for security analysis. 
            Tool Poisoning Attack detection prevents malicious code execution.
 
  
 
4
 
Optimized Execution
 

Tool calls are routed to appropriate upstream servers with response truncation 
            and caching to prevent LLM context bloat.
 
 
 
 
  
 
 
 

Documentation
 

Dive deeper into the features and configuration of MCPProxy with our comprehensive guides.
  
Visit Full Documentation →
 
 
  
Getting Started
 

Learn how to run `mcpproxy` with command-line arguments and environment variables. Covers data directories and logging.
   
Configuration
 

A detailed look at the mcp_config.json file. Manage upstream servers, including adding, updating, and removing them.
   
Features Deep Dive
 

Understand the core concepts of intelligent tool discovery, token optimization, and the security quarantine system.
  
 
 
  
 
 
 
 

Why MCPProxy Saves Tokens
 
 
 
  
 
 
Single Tool Schema
 

Instead of loading hundreds of function schemas, AI agents see only MCPProxy's
retrieve_tools function.
 
 
 
 
  
 
 
On-Demand Discovery
 

Tools are discovered only when needed, keeping LLM context minimal and focused. 
                  No unused function schemas cluttering the prompt.
 
 
 
 
  
 
 
Response Truncation
 

Large tool responses are automatically truncated with caching for pagination, 
                  preventing context window overflow while maintaining data access.
 
 
 
 
 
 
Research-Backed Results
 
 
 
~99%
 
Token reduction vs. loading all schemas
 
 
 
43%
 
Accuracy improvement over baseline
 
 
 
4,400+
 
MCP servers available on mcp.so
 
 
 

Based on RAG-MCP
and ScaleMCP research.
 
 
 
 
  
 
 
 

Advanced Security Protection
 

MCPProxy implements comprehensive security measures to protect against emerging threats 
          in the MCP ecosystem, including Tool Poisoning Attacks.
 
 
 
 

What are Tool Poisoning Attacks?
 
 

Tool Poisoning Attacks embed malicious instructions within MCP tool descriptions 
              that are invisible to users but visible to AI models. These hidden instructions 
              can manipulate AI models into:
 
 
Accessing sensitive files (SSH keys, configuration files, databases)
 
Exfiltrating data while concealing actions from users
 
Overriding instructions from trusted servers
 
Executing unauthorized system commands
 
 
 
 
 

MCPProxy's Protection
 
 
 
Automatic Quarantine
 

All newly added servers are automatically quarantined for security review.
 
 
 
Security Analysis
 

Comprehensive tool description analysis with TPA pattern detection.
 
 
 
Manual Approval
 

Human review required before servers can execute tools.
 
 
 
Execution Blocking
 

Quarantined servers cannot execute tools until approved.