Smartsheet built a remote Model Context Protocol (MCP) server on Amazon Web Services (AWS), according to a blog post published by AWS. The deployment lets Claude AI agents query and update Smartsheet sheets via standardized tool calls, eliminating local MCP client dependencies.
Key facts
- Smartsheet built a remote MCP server on AWS ECS Fargate.
- Server exposes Smartsheet REST APIs as MCP tools for Claude.
- Authentication uses OAuth 2.0 with a custom authorization service.
- Deployment eliminates local MCP client on developer machines.
- Remote MCP enables centralized access control and rate limiting.
Smartsheet, the work management platform, has deployed a remote Model Context Protocol (MCP) server on AWS, as detailed in a blog post by AWS. The server exposes Smartsheet's REST API endpoints as tools that Claude can invoke, enabling AI agents to read, create, update, and delete rows, sheets, reports, and dashboards programmatically.
How the architecture works
The MCP server runs on Amazon ECS Fargate behind an Application Load Balancer. Authentication uses OAuth 2.0 with a custom authorization service, according to the AWS post. The deployment eliminates the need for a local MCP client on the developer's machine — a key difference from the typical local MCP setup where the client runs as a sidecar process alongside the AI application.
By hosting the MCP server remotely, Smartsheet can enforce centralized access controls, logging, and rate limiting. The server communicates with Claude via the MCP specification, which Anthropic introduced in November 2024 as an open standard for connecting AI models to external data sources according to our prior reporting on MCP.
Why remote MCP matters
The remote deployment pattern addresses a scaling limitation of MCP. In the default local setup, each developer runs their own MCP client, which creates operational overhead for teams. Centralizing the MCP server on AWS Fargate means Smartsheet can manage a single instance that multiple Claude sessions or agents can call.
The approach also aligns with the growing trend of enterprise MCP deployments. On July 28, 2026, Google added background execution and MCP support to Gemini API Managed Agents [per our history]. Amazon launched an MCP server for its Registry of Open Data on the same date. The Smartsheet deployment on AWS signals that enterprises are moving MCP from developer-local experiments to production infrastructure.
Smartsheet did not disclose usage metrics or the number of agents currently using the server. The company also did not specify which Claude model variant the MCP server connects to — Claude 3.5 Sonnet, Claude 4 Opus, or another — though the MCP protocol itself is model-agnostic.
What to watch
Watch for other enterprise SaaS platforms to follow Smartsheet's remote MCP pattern on AWS, GCP, or Azure. The Q3 2026 adoption rate of MCP servers in production (vs. local dev setups) will indicate whether the protocol becomes a standard enterprise integration layer or remains a niche developer tool.
Source: news.google.com
[Updated 19 Jul via devto_mcp]
A developer-built cloud gateway now automates MCP server creation from any OpenAPI spec, eliminating the boilerplate Smartsheet had to write for its custom deployment. The gateway, hosted on Render, parses JSON specifications and provisions a remote MCP server over Server-Sent Events (SSE) instantly — no manual SDK code required. Its architecture uses FastAPI for core logic and Supabase for metadata indexing [per dev.to]. This could lower the barrier for other SaaS platforms to replicate Smartsheet's remote MCP pattern without building custom authorization services.
