MCP Router
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@MCP Routerlist available tools"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP Router
A cloud-agnostic Node.js orchestration server that routes requests to pluggable tool adapters (Claude, Cursor, custom tools) via the MCP protocol pattern. Deployable on EKS (AWS) or AKS (Azure) without code changes.
Architecture decisions
Decision | Choice | Reason |
Protocol | HTTP + SSE | Universal, no WS infra needed, works through all proxies |
Adapter loading | Dynamic | Drop-in plugins, zero router changes |
Auth | Delegated to Ingress | Keeps router stateless and horizontally scalable |
K8s packaging | Helm + plain YAML | Works with ArgoCD, Flux, or raw kubectl |
Image | Distroless nonroot | Minimal attack surface, no shell |
Related MCP server: MCP Director
Quick start
cp .env.example .env # add your ANTHROPIC_API_KEY
npm install
npm run devCall the router
# List available tools
curl http://localhost:3000/mcp/tools
# Stream a Claude completion (SSE)
curl -N -X POST http://localhost:3000/mcp \
-H 'Content-Type: application/json' \
-d '{
"tool": "claude",
"stream": true,
"input": {
"messages": [{"role":"user","content":"Explain MCP in one paragraph"}]
}
}'
# Buffered JSON response
curl -X POST http://localhost:3000/mcp \
-H 'Content-Type: application/json' \
-d '{"tool":"custom","stream":false,"input":{"input":"hello world"}}'Adding a new adapter
cp src/adapters/custom.js src/adapters/my-tool.jsChange
nameinmeta()Update
InputSchemaand implementexecute()as an async generatorRestart — the registry auto-discovers the new file
No changes to the router, registry, or Kubernetes manifests.
Deploy to Kubernetes
Prerequisites
kubectlconfigured for your cluster (EKS or AKS)helmv3Your image pushed to ECR or ACR
Create secrets
kubectl create secret generic mcp-router-secrets \
--from-literal=ANTHROPIC_API_KEY=sk-... \
--from-literal=CURSOR_API_KEY=optionalHelm install
helm install mcp-router ./helm/mcp-router \
--set image.repository=<your-registry>/mcp-router \
--set image.tag=<git-sha>Or plain kubectl
kubectl apply -f k8s/manifests.yamlCloud-specific ingress notes
EKS: install aws-load-balancer-controller and add annotation:
kubernetes.io/ingress.class: alb
alb.ingress.kubernetes.io/scheme: internet-facingAKS: install ingress-nginx via Helm (same chart works on both clouds):
helm install ingress-nginx ingress-nginx/ingress-nginxThe default values.yaml targets nginx, which is cloud-agnostic.
Environment variables
Variable | Required | Description |
| For claude adapter | Anthropic API key |
| For cursor adapter | Cursor REST endpoint URL |
| For cursor adapter | Bearer token for Cursor API |
| For cursor adapter | Path to Cursor binary (alternative to REST) |
| No (default 3000) | HTTP port |
| No (default info) | Pino log level |
| No | Set to |
API reference
POST /mcp
{
"tool": "claude",
"stream": true,
"input": { /* adapter-specific, validated by Zod schema */ }
}SSE events (when stream: true):
data: {"type":"text","data":"Hello"}
data: {"type":"done","data":{"stop_reason":"end_turn","usage":{...}}}
data: {"type":"error","data":"Something went wrong"}GET /mcp/tools
Returns registered adapters and their descriptions.
GET /health/live
Kubernetes liveness probe.
GET /health/ready
Kubernetes readiness probe — 503 until adapters are loaded.
This server cannot be deployed
Maintenance
Related MCP Connectors
MCP server for AI agents to plan, verify, and deploy Cloudflare-native apps.
MCP server for progressive tool usage at any scale (see https://klavis.ai)
The Remote MCP server acts as a standardized bridge between LLM applications (like Claude, ChatGPT, and Cursor) and external services, enabling AI agents to access external tools and resources. Its primary capability is providing a centralized search tool to discover other MCP servers and their respective tools. Unlike local implementations, it runs remotely with OAuth authentication and permission controls for security.
MCP server connecting AI agents to 100+ apps (Gmail, Slack, Notion, GitHub) via one-click OAuth.
Related MCP Servers
- FlicenseAqualityDmaintenanceAn intelligent MCP server that orchestrates multiple MCP servers with AI-enhanced workflow automation and production-ready context engine capabilities for codebase analysis.37-
- -licenseNot gradedqualityNot gradedmaintenanceAn intelligent orchestration server that routes user requests to appropriate specialized MCP servers and manages complex workflows.-
- AlicenseBqualityDmaintenanceAn MCP server that enables Claude to manage infrastructure across Kubernetes, Docker, Prometheus, and Terraform through natural language. It provides over 42 specialized tools with a safety-first design, including risk-based command classification and audit logging.43MIT
- AlicenseAqualityCmaintenanceA single MCP server that fronts multiple REST APIs, each configured via environment variables, allowing Claude to orchestrate across several SaaS backends with namespaced tools.21MIT