​
Overview

MCP Agent Cloud deploys your agents as MCP servers, where agent workflows become long-running MCP tools executed in a durable workflow orchestration engine (Temporal). This architecture enables:
  • Agents as MCP Servers: Your agents are exposed as standard MCP servers that any MCP client can connect to
  • Workflows as Durable Tools: Agent workflows are exposed as MCP tools that run durably via Temporal
  • Production-Ready Infrastructure: Built-in retry logic, fault tolerance, and state persistence

​
Key Features

  • Agent Deployment: Deploy agents as MCP servers accessible via HTTP/WebSocket
  • Temporal Integration: Durable workflow execution with automatic retries
  • Secrets Management: Secure storage and injection of API keys
  • Multi-Region Support: Deploy to different regions for lower latency
  • Monitoring: Real-time logs, metrics, and workflow tracking

​
Deployment Process

  1. Write your agent using mcp-agent framework
  2. Configure deployment settings
  3. Deploy with mcp-agent deploy
  4. Access your agent via MCP protocol

​
Architecture

​
Core Concept: Agents as MCP Servers

When you deploy an agent to MCP Agent Cloud:
  1. Your agent becomes an MCP server - Accessible via standard MCP protocol
  2. Workflows become MCP tools - Each workflow is exposed as a tool that MCP clients can invoke
  3. Execution is durable - Workflows run on Temporal, surviving failures and restarts
  4. Tools are long-running - Unlike traditional tools, agent workflows can run for hours/days/weeks

​
How It Works

# Your agent definition
@app.workflow
class ResearchWorkflow:
    async def run(self, topic: str):
        # Long-running research task
        return results

# Becomes an MCP tool when deployed:
# Tool: "workflows-ResearchWorkflow-run"
# Returns: workflow_id and run_id for tracking
# Status: Check via "workflows-get_status" 
# Cancel: Use "workflows-cancel"

​
Components

  • API Gateway: Exposes agents as MCP servers (SSE/HTTP endpoints)
  • Agent Runtime: Containerized execution environment for your agent code
  • Temporal Cluster: Durable workflow orchestration engine
    • Workflows survive crashes and restarts
    • Automatic retries with exponential backoff
    • Pause/resume capabilities for human-in-the-loop
    • Complete execution history and replay debugging
  • Vault: Secure secrets management
  • Monitoring Stack: OpenTelemetry-based observability

​
Why Temporal for Agent Workflows?

Agent workflows are fundamentally different from simple tool calls:
  • They can run for extended periods (hours/days)
  • They need to survive infrastructure failures
  • They require pause/resume for human approval
  • They benefit from automatic retries and error handling
Temporal provides all these capabilities out-of-the-box, making your agents production-ready.

​
Platform Capabilities

​
Deployment

  • One-command deployment via CLI
  • Automatic versioning
  • Blue-green deployments
  • Regional deployment options
  • Custom domains

​
Workflow Management

  • Start, pause, resume, cancel operations
  • Cron-based scheduling
  • Batch operations
  • Priority queues
  • Rate limiting

​
Observability

  • Real-time workflow monitoring
  • Structured logging
  • OpenTelemetry tracing
  • Custom metrics
  • Workflow health checks

​
Security

  • HashiCorp Vault integration
  • End-to-end encryption
  • Role-based access control
  • Audit logging

​
Integrations

  • MCP client compatibility (Claude Desktop, VS Code)
  • Webhook notifications
  • GitHub Actions
  • Custom MCP servers
  • Monitoring exports (Datadog, Grafana)

​
Use Cases

  • Customer Support: Deploy agents for handling support tickets
  • Data Analysis: Process and analyze business data
  • Documentation: Generate and maintain documentation
  • Code Review: Automate code review processes
  • Multi-Agent Systems: Coordinate multiple specialized agents
  • RAG Applications: Retrieval-augmented generation workflows
  • Automation: Business process automation
  • Integration: Connect AI to existing systems

​
Technical Specifications

  • Protocol: Native MCP (Model Context Protocol)
  • Orchestration: Temporal workflow engine
  • Secrets: HashiCorp Vault
  • Deployment: Containerized with Kubernetes
  • Monitoring: OpenTelemetry-based
  • Languages: Python (primary), JavaScript/TypeScript support planned
  • Regions: US East, US West, EU West (more coming)

​
Getting Started

​
Prerequisites

  1. mcp-agent installed (uv tool install mcp-agent)
  2. MCP Agent Cloud account
  3. API keys configured

​
Quick Start

# Install CLI
uv tool install mcp-agent

# Configure credentials
mcp-agent configure

# Deploy your agent
mcp-agent deploy --app my_agent:app

​
Next Steps