mcp -ai

MCP Server AI is a unified, high-performance microservice that abstracts access to multiple IA providers (such as AWS Bedrock and Azure OpenAI) behinda single API surface. It exposes HTTP/REST, gRPC, WebSocket, and server-sent events, enabling multi-protocol clients to interact with a single backend while benefiting from session management, worker pools, caching, and observability. You can reach the API through the HTTP port for REST endpoints, connect via gRPC for high-throughput calls, or use WebSocket for real-time bi-directional communication. The server also supports batch processing and provides monitoring through Prometheus/Grafana, with structured logging and distributed tracing for troubleshooting and performance analysis.

Prerequisites:

• Docker and Docker Compose installed
• Optional: Kubernetes (kubectl) for deployments
• Environment with access to AWS/Azure IA credentials

1. Clone the repository or pull the Docker image

• If using the provided Docker image, you can pull it from Docker Hub (example shown below)

2. Quickstart with Docker Compose (recommended)

• Create or edit a docker-compose.yml to include Redis, PostgreSQL, and the MCP Server AI service. Then start the stack:

# Example docker-compose up the stack (adjust services as needed)
docker-compose up -d

3. Run directly with Docker (alternative)

• Start the MCP server container (example usage):

docker run -d --name mcp-server-ai \
  -p 8090:8090 -p 8091:8091 -p 50051:50051 \
  -e SERVER_ENV=production \
  proyectoskevinsvega/mcp-server-ai:latest

4. Optional: Kubernetes deployment

• Use the provided Helm charts or Kubernetes manifests in deploy/k8s to deploy with HPA/VPA and monitoring. Ensure secrets are created for IA credentials before deploying.

5. Environment configuration

• You can configure via environment variables (see README variables section) or a .env file mounted into the container. Typical variables include provider credentials, Redis/PostgreSQL connection strings, and service ports.

Tips and common issues:

• Ensure your IA provider credentials (AWS/Azure) are valid and have the required permissions.
• When using Redis, enable TLS and clustering in production for improved resilience.
• If you see high latency, check worker pool scaling settings and adjust MAX_TOKENS, TEMPERATURE, and the number of workers accordingly.
• For production deployments, enable RBAC, Secrets management (Vault/Sealed Secrets), and implement network policies for Kubernetes.
• Review the environment variables for CORS and API ports to avoid conflicts in multi-tenant environments.
• Use Prometheus/Grafana dashboards to monitor latency, throughput, and resource utilization; enable Jaeger tracing for distributed traces across services.