Tributary-ai-services/tas-mcp

📡 TAS Model Context Protocol (MCP) Server

The TAS MCP Server is a high-performance, cloud-native event gateway and ingestion service that implements the Model Context Protocol to support RAG pipelines, event-driven architectures, and workflow orchestration across distributed AI systems.

🌟 Key Features

• 🚀 Multi-Protocol Support: HTTP REST API and bidirectional gRPC streaming
• 🔄 Smart Event Forwarding: Rule-based routing with condition evaluation
• 🎯 Event Transformation: Template-based and programmatic event transformation
• 🔌 Integration Ready: Native support for Argo Events, Kafka, webhooks, and more
• 📚 MCP Server Registry: Comprehensive catalog of MCP servers and capabilities
• 📊 Observability: Built-in metrics, health checks, and distributed tracing support
• 🔒 Production Ready: Rate limiting, circuit breakers, and retry logic
• ☁️ Cloud Native: Kubernetes-native with Helm charts and operators
• 🎨 Extensible: Plugin architecture for custom forwarders and processors

🚀 Quick Start

Using Docker

# Run with Docker
docker run -p 8080:8080 -p 50051:50051 ghcr.io/tributary-ai-services/tas-mcp:latest

# Or build locally
make docker
make docker-run

Using Docker Compose

# Start all services
make docker-compose

# View logs
docker-compose logs -f tas-mcp-server

Local Development

# Install dependencies
make init

# Run locally
make run

# Run with hot reload
make dev

📡 API Usage

HTTP API

# Ingest a single event
curl -X POST http://localhost:8080/api/v1/events \
  -H "Content-Type: application/json" \
  -d '{
    "event_id": "evt-123",
    "event_type": "user.created",
    "source": "auth-service",
    "data": "{\"user_id\": \"usr-456\", \"email\": \"user@example.com\"}"
  }'

# Batch event ingestion
curl -X POST http://localhost:8080/api/v1/events/batch \
  -H "Content-Type: application/json" \
  -d '[
    {"event_id": "evt-1", "event_type": "order.created", "source": "order-service", "data": "{}"},
    {"event_id": "evt-2", "event_type": "payment.processed", "source": "payment-service", "data": "{}"}
  ]'

# Health check
curl http://localhost:8082/health

gRPC API

// Example Go client
import (
    mcpv1 "github.com/tributary-ai-services/tas-mcp/gen/mcp/v1"
    "google.golang.org/grpc"
)

conn, _ := grpc.Dial("localhost:50051", grpc.WithInsecure())
client := mcpv1.NewMCPServiceClient(conn)

// Ingest event
resp, _ := client.IngestEvent(ctx, &mcpv1.IngestEventRequest{
    EventId:   "evt-123",
    EventType: "user.action",
    Source:    "webapp",
    Data:      `{"action": "login"}`,
})

// Stream events
stream, _ := client.EventStream(ctx)

🔧 Configuration

Environment Variables

Configuration File

{
  "HTTPPort": 8080,
  "GRPCPort": 50051,
  "LogLevel": "info",
  "forwarding": {
    "enabled": true,
    "targets": [
      {
        "id": "argo-events",
        "type": "webhook",
        "endpoint": "http://argo-events-webhook:12000",
        "rules": [
          {
            "conditions": [
              {"field": "event_type", "operator": "contains", "value": "critical"}
            ]
          }
        ]
      }
    ]
  }
}

🏗️ Architecture

The TAS MCP Server follows a modular, event-driven architecture:

┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│   HTTP API  │     │  gRPC API   │     │   Webhook   │
└──────┬──────┘     └──────┬──────┘     └──────┬──────┘
       │                   │                    │
       └───────────────────┴────────────────────┘
                           │
                    ┌──────▼──────┐
                    │   Ingestion │
                    │    Layer    │
                    └──────┬──────┘
                           │
                    ┌──────▼──────┐
                    │    Rules    │
                    │   Engine    │
                    └──────┬──────┘
                           │
       ┌───────────────────┴────────────────────┐
       │                                        │
┌──────▼──────┐  ┌──────────────┐  ┌──────────▼────┐
│  Forwarder  │  │  Transform   │  │   Metrics     │
│   (gRPC)    │  │   Engine     │  │  Collector    │
└─────────────┘  └──────────────┘  └───────────────┘

🚢 Deployment

Kubernetes

# Deploy to development
kubectl apply -k k8s/overlays/dev

# Deploy to production with auto-scaling
kubectl apply -k k8s/overlays/prod

# Check deployment status
kubectl get pods,svc,ingress -n tas-mcp-prod

Helm Chart (Coming Soon)

helm repo add tas-mcp https://tributary-ai-services.github.io/helm-charts
helm install tas-mcp tas-mcp/tas-mcp-server

📚 MCP Server Registry

The TAS MCP project includes a comprehensive MCP Server Registry - a curated catalog of Model Context Protocol servers across different categories and use cases.

🗂️ Registry Categories

• 🤖 AI Models - LLM integrations and model serving
• 📡 Event Streaming - Real-time event processing and forwarding
• 🔄 Workflow Orchestration - Complex workflow and agent coordination
• 💾 Knowledge Bases - Vector stores and search capabilities
• 🔧 Data Processing - ETL and data transformation services
• 📊 Monitoring - Observability and metrics collection
• 💬 Communication - Chat bots and messaging integrations
• 🔍 Search - Web search and content discovery services
• 🕷️ Web Scraping - Data extraction and automation tools
• 🗃️ Database - Database integration and query services
• 🛠️ Development Tools - Git, CI/CD, and development utilities

🚀 Using the Registry

# Browse the registry
cat registry/mcp-servers.json | jq '.servers[] | select(.category == "ai-models")'

# Find servers by capability
cat registry/mcp-servers.json | jq '.servers[] | select(.capabilities[] | contains("event-streaming"))'

# Get deployment information
cat registry/mcp-servers.json | jq '.servers[] | select(.name == "tas-mcp-server") | .deployment'

# Find privacy-focused search servers
cat registry/mcp-servers.json | jq '.servers[] | select(.category == "search" and .privacy.noTracking == true)'

# Find web scraping servers
cat registry/mcp-servers.json | jq '.servers[] | select(.category == "web-scraping")'

# Find database integration servers
cat registry/mcp-servers.json | jq '.servers[] | select(.category == "database")'

📋 Registry Features

• JSON Schema Validation - Ensures data consistency and structure
• Deployment Ready - Docker and Kubernetes deployment configurations
• Access Models - Clear documentation of API access patterns
• Capability Mapping - Searchable capability tags
• Cost Information - Pricing and resource requirements

See for complete registry documentation and for integration guides.

🔌 Integrations

TAS MCP Federation Servers

The project includes several fully-integrated MCP servers ready for deployment:

🔍 DuckDuckGo MCP Server

• Privacy-focused web search with no tracking or data collection
• News search with time filtering capabilities
• Image search with advanced filters (size, color, type)
• Content extraction from web pages
• Deployment: deployments/docker-compose/duckduckgo-mcp/

🕷️ Apify MCP Server

• Access to 5,000+ web scraping actors from the Apify platform
• E-commerce, social media, and news scraping
• Custom scraping configurations and data extraction
• Dataset management and export capabilities
• Deployment: deployments/docker-compose/apify-mcp/

🗃️ PostgreSQL MCP Server

• Read-only database access with security-first design
• Schema inspection and table metadata
• Query execution with performance analysis
• Connection pooling and health monitoring
• Deployment: deployments/postgres-mcp/

🛠️ Git MCP Server

• Repository interaction and automation
• Branch management and commit operations
• Status and diff operations
• Working tree management
• Based on official Model Context Protocol Git server

Full-Stack Deployment

# Deploy complete federation stack
cd deployments/docker-compose
docker-compose -f full-stack.yml up -d

# Check all services
docker-compose -f full-stack.yml ps

# View federation status
curl http://localhost:8080/api/v1/federation/servers | jq '.'

See for complete integration examples in Go.

Argo Events

See for complete integration examples in Go, Python, and Node.js.

Kafka

{
  "type": "kafka",
  "endpoint": "kafka-broker:9092",
  "config": {
    "topic": "mcp-events",
    "batch_size": 100
  }
}

Prometheus Metrics

The server exposes Prometheus metrics at /api/v1/metrics:

• mcp_events_total - Total events processed
• mcp_events_forwarded_total - Events forwarded by target
• mcp_forwarding_errors_total - Forwarding errors by target
• mcp_event_processing_duration_seconds - Event processing latency

🧪 Testing

The project includes comprehensive test coverage across all packages:

# Run all unit tests
make test-unit

# Run integration tests
make test-integration

# Run benchmark tests
make test-benchmark

# Generate coverage report
make test-coverage

# Run all tests (unit + integration + benchmarks)
make test

# Lint code
make lint

# Format code
make fmt

Test Coverage

• Config Package: 77.6% statement coverage
• Forwarding Package: 60.1% statement coverage
• Integration Tests: End-to-end event forwarding scenarios
• Benchmark Tests: Performance testing for critical paths

Test Features

• Table-driven tests for comprehensive scenario coverage
• Mock HTTP servers for integration testing
• Event matching and validation utilities
• Concurrent testing patterns
• Test utilities package for reusable helpers

🗺️ Roadmap

See our comprehensive for detailed development priorities, including:

• 1,535+ MCP server federation across 12 categories from mcpservers.org
• Universal MCP Orchestrator vision and implementation plan
• Quarterly release schedule with progressive federation milestones
• Community involvement opportunities and feedback channels
• Technical implementation plans for massive-scale federation

🤝 Contributing

We welcome contributions! Please see for guidelines.

Development Setup

See for detailed development instructions.

# Setup development environment
make init

# Run tests and linting
make test lint

# Submit changes
make fmt
git add .
git commit -m "feat: add new feature"

📚 Documentation

• - Development setup and guidelines
• - Complete API documentation
• - System design and architecture
• - Container deployment guide
• - Integration examples and tutorials

🔐 Security

• Non-root container execution
• TLS support for all protocols
• Authentication via API keys or OAuth2
• Rate limiting and DDoS protection
• Regular security scanning with Trivy

Report security vulnerabilities to: security@tributary-ai-services.com

📊 Performance

• Handles 10,000+ events/second per instance
• Sub-millisecond forwarding latency
• Horizontal scaling with Kubernetes HPA
• Efficient memory usage with bounded buffers
• Connection pooling for downstream services

🗺️ Roadmap

• Helm chart for easy deployment
• WebSocket support for real-time streaming
• Event replay and time-travel debugging
• GraphQL API for flexible queries
• Built-in event store with retention policies
• SDK libraries for popular languages
• Terraform modules for cloud deployment

📄 License

This project is licensed under the Apache License 2.0 - see the file for details.

🙏 Acknowledgments

• Model Context Protocol by Anthropic
• Argo Events for event-driven workflows
• The Go community for excellent libraries and tools

Built with ❤️ by Tributary AI Services