How Eon built multi-cloud investigation with an MCP Gateway on Amazon Bedrock AgentCore
Eon replaced per‑tool credentials and local Model Context Protocol (MCP) servers with a single authenticated MCP Gateway built on Amazon Bedrock AgentCore, cutting new‑engineer onboarding from roughly two days to about two minutes. The unified gateway also lets their Claude‑based autonomous SRE agent access the same multi‑cloud investigation tools with role‑based permissions. By storing investigation data in Apache Parquet and Apache Iceberg under a single read‑only AWS IAM role, Eon measured up to a 50 % reduction in storage costs. Engineers can now provision cross‑cloud tool access without VPC peering or PrivateLink, dramatically simplifying credential management and scaling.
⚡ Key Takeaways
- Onboarding time dropped from ~2 days to ~2 minutes for new engineers.
- Centralized MCP Gateway runs on Amazon Bedrock AgentCore, exposing a single authenticated entry point for AWS, GCP, and Azure tools.
- Credential consolidation eliminates the need for per‑tool auth configs and local MCP servers, reducing operational overhead.
- The solution stores data in Apache Parquet and Apache Iceberg formats using a single read‑only AWS IAM role.
- Cross‑account access is achieved without VPC peering or AWS PrivateLink, lowering networking cost and setup complexity.
Engineers can provision secure, low‑latency access to dozens of multi‑cloud operational tools instantly, enabling faster incident response and cost‑effective data retention for production AI systems.
✅ Practical Steps
- Deploy an MCP Gateway on Amazon Bedrock AgentCore as described, configuring it to route requests to AWS, GCP, and Azure investigation tools.
- Assign a read‑only AWS IAM role to the gateway and configure data pipelines to write investigation artifacts to Apache Parquet and Apache Iceberg.
- Update both human developers and the Claude‑based autonomous SRE agent to authenticate through the single gateway endpoint instead of individual tool credentials.
Want the full story? Read the original article.
Read on AWS ML Blog ↗