AWS introduces multi-account AI agent architecture with AgentCore Gateway and MCP
AWS has detailed a new architecture for enterprises to build AI agents that query data across multiple AWS accounts without centralizing it. The solution uses AgentCore Gateway and Model Context Protocol (MCP) to let each line-of-business (LOB) team keep its data isolated while agents access only the required information at query time. The platform account hosts the agent and inference through…
Key points
- AgentCore Gateway aggregates MCP servers from multiple AWS accounts into a single endpoint for unified tool discovery and invocation
- LOB teams expose data as MCP servers on AgentCore Runtime, keeping ownership and isolation while enabling agent access
- Security relies on OAuth 2.0 M2M credentials, Okta/OIDC validation, and Policy in AgentCore for fine-grained authorization
The setup avoids replicating data or untangling cross-account IAM complexities. Each LOB’s MCP server runs on AgentCore Runtime, a serverless environment with session isolation and consumption-based pricing. The platform team controls models, applies guardrails, and tracks costs centrally, while LOB teams retain ownership of their tool surfaces. Security measures include OAuth 2.0 machine-to-machine (M2M) credentials, Okta integration, and allowedWorkloadConfiguration to restrict direct access. Observability tools like AgentCore Evaluations and Amazon CloudWatch monitor performance, and cost attribution tags help allocate expenses back to the originating LOB. The architecture scales by adding new LOB targets to the Gateway, with no changes required to the agent’s code.
Build a multi-account AI agent with AgentCore Gateway and MCP
AWS Machine Learning Blog · 24 September 2026
Loading the full article…
This text was published by AWS Machine Learning Blog and written by Senthil Kamala Rathinam. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Enterprise & Industry
All →- Gemini introduces 3.8 Live with Live Avatar · 2 src
- Oracle sends force majeure notice for New Mexico Stargate data center · 1 src
- BNP Paribas signs five-year agentic AI deal with Google Cloud · 1 src
- Anthropic partners with OpenEvidence to bring free medical AI to 100 countries · 2 src
- Trebellar raises $18M Series A to add AI agents to corporate real estate · 1 src
Comments
via GitHub Discussions