Smartsheet Builds AI-Optimized MCP Server on AWS for Enterprise Data Access

Smartsheet, a leading enterprise work management platform, has implemented a remote MCP server on AWS to facilitate structured AI access to its data and capabilities. This architecture allows both internal AI features, like Smart Assist, and external AI clients, such as Amazon Quick and Claude Desktop, to interact seamlessly with Smartsheet's data. The MCP server connects to existing APIs and an intelligence layer built on Amazon Neptune and Amazon S3, optimized for AI workloads.

The architecture employs AWS services like Fargate for containerized server deployment, Kinesis Data Streams for event ingestion, and Bedrock for large language model inference. It supports bursty AI traffic through auto-scaling policies and ensures smooth updates via CI/CD pipelines with automated rollback mechanisms. Security is reinforced with AWS WAF, Shield, mutual TLS, and OAuth2, while layered rate limiting manages AI-specific traffic patterns.

Smartsheet emphasizes governance and observability, integrating access controls, audit trails, and detailed telemetry via Amazon OpenSearch and CloudWatch. The system also incorporates advanced token management strategies, including proprietary serialization formats and schema-driven tool contracts, to reduce costs and prevent hallucinations in AI responses.

This infrastructure enables enterprise clients to deploy autonomous AI agents that perform tasks like requirements gathering, task management, and documentation, all within existing workflows. Smartsheet plans to evolve this platform further with AWS services like Amazon Bedrock AgentCore, supporting personalized, autonomous, and scalable AI solutions.

The development of this AWS-based AI infrastructure positions Smartsheet as a leader in enterprise AI integration, offering scalable, secure, and cost-efficient solutions that enhance productivity and data governance for large organizations.