svrnty-mcp-gateway/docs/module-design.md

Svrnty.MCP.Gateway - Module Design

Document Type: Architecture Design Document Status: Planned Version: 1.0.0 Last Updated: 2025-10-19

---

Overview

Svrnty.MCP.Gateway is a .NET 8 library that provides proxy and routing infrastructure for MCP traffic, enabling centralized management, authentication, monitoring, and load balancing between MCP clients and servers. This document defines the architecture, components, and design decisions.

Purpose

• What: Gateway/proxy library for routing MCP traffic between clients and servers
• Why: Enable centralized management, security, and monitoring of MCP infrastructure
• How: Clean Architecture with routing strategies, health monitoring, and transport abstraction

---

Architecture

Clean Architecture Layers

┌─────────────────────────────────────────────────┐
│      Svrnty.MCP.Gateway.Cli (Executable)    │
│  ┌───────────────────────────────────────────┐  │
│  │  Svrnty.MCP.Gateway.AspNetCore (HTTP)│  │
│  │  ┌─────────────────────────────────────┐ │  │
│  │  │ Svrnty.MCP.Gateway.Infrastructure│ │  │
│  │  │  ┌───────────────────────────────┐  │ │  │
│  │  │  │ Svrnty.MCP.Gateway.Core   │  │ │  │
│  │  │  │  - IGatewayRouter             │  │ │  │
│  │  │  │  - IRoutingStrategy           │  │ │  │
│  │  │  │  - IAuthProvider              │  │ │  │
│  │  │  │  - ICircuitBreaker            │  │ │  │
│  │  │  │  - Models (no dependencies)   │  │ │  │
│  │  │  └───────────────────────────────┘  │ │  │
│  │  └─────────────────────────────────────┘ │  │
│  └───────────────────────────────────────────┘  │
└─────────────────────────────────────────────────┘

Layer Responsibilities

Layer	Purpose	Dependencies
Core	Abstractions and models	None
Infrastructure	Routing, auth, circuit breakers	Core, System.Text.Json
AspNetCore	HTTP endpoints and DI	Core, Infrastructure, ASP.NET Core
Cli	Management CLI	All layers

---

Core Components

IGatewayRouter Interface

Primary interface for gateway routing operations:

public interface IGatewayRouter
{
    // Request Routing
    Task<McpResponse> RouteRequestAsync(
        McpRequest request,
        RoutingContext context,
        CancellationToken ct = default
    );

    // Server Management
    Task<IEnumerable<ServerInfo>> GetRegisteredServersAsync();
    Task RegisterServerAsync(ServerConfig config);
    Task UnregisterServerAsync(string serverId);

    // Health Monitoring
    Task<ServerHealthStatus> GetServerHealthAsync(string serverId);
    Task<IEnumerable<ServerHealthStatus>> GetAllServerHealthAsync();
}

IRoutingStrategy Interface

Defines server selection logic:

public interface IRoutingStrategy
{
    string SelectServer(
        RoutingContext context,
        IEnumerable<ServerInfo> availableServers
    );
}

public class RoutingContext
{
    public string? ToolName { get; set; }
    public string? ClientId { get; set; }
    public Dictionary<string, string>? Headers { get; set; }
    public Dictionary<string, object>? Metadata { get; set; }
}

IAuthProvider Interface

Authentication and authorization:

public interface IAuthProvider
{
    Task<AuthResult> AuthenticateAsync(
        string? apiKey,
        Dictionary<string, string>? headers,
        CancellationToken ct = default
    );

    Task<bool> AuthorizeAsync(
        string clientId,
        string serverId,
        string toolName,
        CancellationToken ct = default
    );
}

public class AuthResult
{
    public bool IsAuthenticated { get; set; }
    public string? ClientId { get; set; }
    public IEnumerable<string> Roles { get; set; } = [];
    public string? ErrorMessage { get; set; }
}

ICircuitBreaker Interface

Prevent cascading failures:

public interface ICircuitBreaker
{
    bool IsOpen(string serverId);
    void RecordSuccess(string serverId);
    void RecordFailure(string serverId);
    void Reset(string serverId);
}

---

Routing Strategies

Built-In Strategies

Round-Robin Strategy

public class RoundRobinStrategy : IRoutingStrategy
{
    private int _currentIndex = 0;

    public string SelectServer(
        RoutingContext context,
        IEnumerable<ServerInfo> servers)
    {
        var healthyServers = servers.Where(s => s.IsHealthy).ToList();

        if (healthyServers.Count == 0)
        {
            throw new NoHealthyServersException();
        }

        var index = Interlocked.Increment(ref _currentIndex) % healthyServers.Count;
        return healthyServers[index].Id;
    }
}

Tool-Based Strategy

public class ToolBasedStrategy : IRoutingStrategy
{
    private readonly Dictionary<string, string> _toolPrefixMappings;

    public string SelectServer(
        RoutingContext context,
        IEnumerable<ServerInfo> servers)
    {
        if (context.ToolName == null)
        {
            throw new InvalidOperationException("ToolName required for tool-based routing");
        }

        foreach (var (prefix, serverId) in _toolPrefixMappings)
        {
            if (context.ToolName.StartsWith(prefix))
            {
                return serverId;
            }
        }

        // Default to first healthy server
        return servers.First(s => s.IsHealthy).Id;
    }
}

Client-Based Strategy

public class ClientBasedStrategy : IRoutingStrategy
{
    private readonly Dictionary<string, string> _clientMappings;

    public string SelectServer(
        RoutingContext context,
        IEnumerable<ServerInfo> servers)
    {
        if (context.ClientId != null &&
            _clientMappings.TryGetValue(context.ClientId, out var serverId))
        {
            return serverId;
        }

        // Default routing
        return servers.First(s => s.IsHealthy).Id;
    }
}

---

Configuration

Gateway Configuration Model

public class GatewayConfig
{
    public string Name { get; set; } = "MCP Gateway";
    public string Version { get; set; } = "1.0.0";
    public string? Description { get; set; }
    public string ListenAddress { get; set; } = "http://localhost:8080";
}

public class ServerConfig
{
    public string Id { get; set; } = string.Empty;
    public string Name { get; set; } = string.Empty;
    public TransportConfig Transport { get; set; } = new();
    public bool Enabled { get; set; } = true;
    public Dictionary<string, string>? Metadata { get; set; }
}

public class TransportConfig
{
    public string Type { get; set; } = "Stdio"; // "Stdio" or "Http"
    public string? Command { get; set; }
    public string[]? Args { get; set; }
    public string? BaseUrl { get; set; }
    public Dictionary<string, string>? Headers { get; set; }
}

Routing Configuration

public class RoutingConfig
{
    public string Strategy { get; set; } = "RoundRobin"; // "RoundRobin", "ToolBased", "ClientBased"
    public TimeSpan HealthCheckInterval { get; set; } = TimeSpan.FromSeconds(30);
    public Dictionary<string, string>? StrategyConfig { get; set; }
}

Security Configuration

public class SecurityConfig
{
    public bool EnableAuthentication { get; set; } = false;
    public string? ApiKeyHeader { get; set; } = "X-MCP-API-Key";
    public RateLimitConfig? RateLimit { get; set; }
}

public class RateLimitConfig
{
    public int RequestsPerMinute { get; set; } = 100;
    public int BurstSize { get; set; } = 20;
}

---

Health Monitoring

Server Health Check

public class ServerHealthStatus
{
    public string ServerId { get; set; } = string.Empty;
    public string ServerName { get; set; } = string.Empty;
    public bool IsHealthy { get; set; }
    public DateTime LastCheck { get; set; }
    public TimeSpan? ResponseTime { get; set; }
    public string? ErrorMessage { get; set; }
}

Health Check Implementation

public class McpServerHealthCheck : IHealthCheck
{
    private readonly IGatewayRouter _router;

    public async Task<HealthCheckResult> CheckHealthAsync(
        HealthCheckContext context,
        CancellationToken ct = default)
    {
        var statuses = await _router.GetAllServerHealthAsync();
        var healthyCount = statuses.Count(s => s.IsHealthy);
        var totalCount = statuses.Count();

        if (healthyCount == totalCount)
        {
            return HealthCheckResult.Healthy(
                $"All {totalCount} servers healthy");
        }
        else if (healthyCount > 0)
        {
            return HealthCheckResult.Degraded(
                $"{healthyCount}/{totalCount} servers healthy");
        }
        else
        {
            return HealthCheckResult.Unhealthy(
                "No healthy servers available");
        }
    }
}

---

Error Handling

Exception Hierarchy

public class GatewayException : Exception { }

public class NoHealthyServersException : GatewayException { }

public class ServerNotFoundException : GatewayException
{
    public string ServerId { get; }
}

public class RoutingException : GatewayException
{
    public RoutingContext Context { get; }
}

public class AuthenticationException : GatewayException { }

public class RateLimitExceededException : GatewayException
{
    public string ClientId { get; }
    public int RequestsPerMinute { get; }
}

Circuit Breaker Implementation

public class CircuitBreaker : ICircuitBreaker
{
    private readonly ConcurrentDictionary<string, CircuitState> _states = new();
    private readonly int _failureThreshold = 5;
    private readonly TimeSpan _timeout = TimeSpan.FromSeconds(30);

    public bool IsOpen(string serverId)
    {
        if (!_states.TryGetValue(serverId, out var state))
        {
            return false;
        }

        if (state.State == CircuitState.Open &&
            DateTime.UtcNow - state.LastFailure > _timeout)
        {
            // Transition to half-open
            state.State = CircuitState.HalfOpen;
        }

        return state.State == CircuitState.Open;
    }

    public void RecordSuccess(string serverId)
    {
        _states.AddOrUpdate(serverId,
            _ => new CircuitState { State = CircuitState.Closed },
            (_, state) => { state.FailureCount = 0; state.State = CircuitState.Closed; return state; });
    }

    public void RecordFailure(string serverId)
    {
        _states.AddOrUpdate(serverId,
            _ => new CircuitState { FailureCount = 1, LastFailure = DateTime.UtcNow },
            (_, state) =>
            {
                state.FailureCount++;
                state.LastFailure = DateTime.UtcNow;
                if (state.FailureCount >= _failureThreshold)
                {
                    state.State = CircuitState.Open;
                }
                return state;
            });
    }

    public void Reset(string serverId)
    {
        _states.TryRemove(serverId, out _);
    }
}

enum CircuitState
{
    Closed,
    Open,
    HalfOpen
}

---

Testing Strategy

Unit Tests

• Test Core abstractions with mocks
• Test routing strategies with mock servers
• Test circuit breaker logic
• Test authentication/authorization

Integration Tests

• Test actual routing to real MCP servers
• Test health checks
• Test error scenarios (server failures, timeouts)
• Test authentication flows

Test Coverage Goals

• Core: >90%
• Infrastructure: >80%
• AspNetCore: >70%

---

Performance Considerations

Connection Pooling

• Maintain persistent connections to backend servers
• Configurable pool size per server
• Idle connection eviction

Request Caching

• Cache tool discovery results
• Cache health check results (with TTL)
• Invalidate cache on server changes

Monitoring

• Track request latency per server
• Track request success/failure rates
• Track circuit breaker state changes
• OpenTelemetry metrics integration

---

Security

Input Validation

• Validate all incoming requests
• Sanitize routing context data
• Validate server configuration

Authentication

• API key authentication
• JWT token support
• Client identity verification

Authorization

• Role-based access control
• Server-level permissions
• Tool-level permissions

Rate Limiting

• Per-client rate limiting
• Per-server rate limiting
• Global rate limiting
• Burst protection

---

Future Enhancements

• WebSocket transport support
• Request/response compression
• Dynamic server registration/discovery
• A/B testing support
• Blue/green deployment routing
• Multi-region routing
• Request replay for debugging
• Distributed tracing integration

---

Document Version: 1.0.0 Status: Planned Next Review: After Phase 1 implementation