Smartsheet Builds Remote MCP Server on AWS for AI Agent Integration with Enterprise Data

Smartsheet, an enterprise work management platform, built a remote MCP server on AWS to provide AI agents with structured access to its data. The server supports both internal Smart Assist and external clients like Amazon Quick and Claude Desktop. It runs on AWS Fargate for containers, uses Amazon Bedrock for LLM inference, Amazon Neptune for cross-project insights, and Amazon Kinesis for event streaming. The server includes progressive disclosure to cap token consumption, strongly typed tool schemas, and a proprietary serialization format that reduces token count by 35-47%. Since launch, Smartsheet saved over 3 billion tokens through these optimizations. The server also features tiered access control, OpenTelemetry-based observability, and layered rate limiting via AWS WAF. Testing includes end-to-end workflows with LLMs in the loop. Deployments follow a canary strategy across AWS regions.