The intelligent MCP server that ends the "Address already in use" nightmare for AI agents working with FastAPI

Every developer knows this pain:

AI Agent: "Let me restart the server..."
> uvicorn main:app --reload --port 8000
[ERROR] Address already in use

AI Agent: "Let me check what's using port 8000..."
AI Agent: "Finding process..."
AI Agent: "Killing process..."
AI Agent: "Starting again..."

[10 minutes and 5000 tokens later...]

You: "The server was already running with --reload 😑"

This happens. Every. Single. Time.

FastDEV is an intelligent MCP (Model Context Protocol) server that acts as a smart intermediary between AI agents and your FastAPI applications. Instead of blindly executing commands, FastDEV understands context and makes intelligent decisions.

Without FastDEV:

# AI wastes 20+ actions on simple operations
agent: "Starting server"
agent: "Error, port in use"  
agent: "Searching for processes"
agent: "Killing process"
agent: "Starting again"
# ... and round and round we go

With FastDEV:

# AI does exactly what's needed
agent: fastdev.ensure_running("my-app")
# FastDEV: "Server my-app already running on :8000 with hot-reload ✓"

Result: 95% fewer tokens, 99% less time, 100% less frustration

• Smart startup - Never launches duplicates, auto-finds free ports
• Context awareness - Knows when hot-reload is active, no unnecessary restarts
• Mode management - Dev, prod, test modes with appropriate settings
• Persistent state - Remembers servers between sessions

agent: fastdev.diagnose_crash("my-app")

# Instead of scrolling through logs:
{
  "crashed_at": "2024-01-15 10:23:45",
  "reason": "ImportError: 'User' from 'models'",
  "solution": "Add 'from models.user import User'",
  "similar_issues": ["#123", "#89"]
}

• Server health status
• Performance metrics
• Memory usage
• Request statistics
• Error tracking

• Multi-tool analysis - Black, isort, mypy, ruff, bandit, flake8, pylint in one call
• AST-based insights - Find dead code, complexity, duplicates
• Auto-fixing - Apply corrections automatically
• Smart suggestions - Context-aware improvements

# Turn FastAPI endpoints into MCP tools
agent: fastdev.bridge_endpoints("my-app")

# AI can now directly call your API
agent: my_app.get_user(user_id=123)
agent: my_app.create_post(title="Hello", content="World")

pip install fastdev-mcp

1. Start FastDEV MCP server:

fastdev serve

2. Configure your AI agent:

For Claude Desktop, add to claude_desktop_config.json:

{
  "mcpServers": {
    "fastdev": {
      "command": "fastdev",
      "args": ["serve"]
    }
  }
}

3. Use in your AI conversations:

# Start a server intelligently
fastdev.start_server("my-app", mode="dev")

# Get logs without drama
fastdev.get_logs("my-app", tail=50)

# Run tests in isolation
fastdev.run_tests("my-app")

# Analyze code quality
fastdev.analyze_code("my-app")

# Check server health
fastdev.health_check("my-app")

# AI Agent uses FastDEV
agent: fastdev.ensure_running("my-api", mode="dev")

# FastDEV responds intelligently:
# ✓ Server "my-api" already running on port 8000 with hot-reload
# ✓ Last reloaded: 30 seconds ago
# ✓ Status: healthy

# When server crashes
agent: fastdev.diagnose_crash("my-api")

# FastDEV provides actionable insights:
{
  "error": "ModuleNotFoundError: pydantic",
  "solution": "pip install pydantic>=2.0",
  "context": "Added in requirements.txt but not installed"
}

# One command instead of running 7 tools
agent: fastdev.analyze_code("my-api")

# FastDEV returns consolidated report:
{
  "critical": ["SQL injection risk in user.py:45"],
  "warnings": ["Unused imports in 3 files"],
  "style": ["Formatting issues in 5 files"],
  "auto_fixable": 23,
  "score": 8.5
}

• Automatic port allocation
• Conflict prevention
• Port persistence across restarts
• Port forwarding configuration

• Graceful shutdowns
• Resource cleanup
• Automatic restart on crash
• Health monitoring

• Hot reload awareness
• Test isolation
• Database migrations
• Environment management

• ✅ Smart server management
• ✅ Automatic port handling
• ✅ Basic crash diagnostics
• ✅ MCP protocol implementation

• 🔄 Full linter integration
• 🔄 AST analysis
• 🔄 Performance profiling
• 🔄 Auto-fixes

• 🔄 FastMCP endpoint bridging
• 🔄 OpenAPI integration
• 🔄 Type validation
• 🔄 Direct API access

• 🔄 Cloud sync
• 🔄 Team collaboration
• 🔄 Plugin marketplace
• 🔄 Enterprise features

• Traditional approach: ~5,000 tokens per "restart" issue
• With FastDEV: ~50 tokens
• Savings: 99%

• Traditional approach: 2-5 minutes of debugging
• With FastDEV: Instant results
• Improvement: 100x faster

• No more port conflicts
• No more unnecessary restarts
• No more token waste
• Just productive development

FastDEV is open source and we love contributions!

• 🐛 Report bugs
• 💡 Suggest features
• 🔧 Submit PRs
• ⭐ Star the project

See CONTRIBUTING.md for details.

MIT License - see LICENSE for details.

• FastAPI for the amazing framework
• Anthropic for the MCP protocol
• FastMCP for the Python MCP framework
• All contributors and early adopters

FastDEV - Because life is too short for "Address already in use"

GitHub • Documentation • Discord • Twitter

Made with ❤️ by developers who are tired of fighting with ports