Amazon Quick now supports VPC connectivity for MCP connections
Enterprise teams can now connect Quick's AI assistant to private, VPC-hosted MCP servers—no public internet exposure required.
View original announcement →Visual Summary
graph TD
A{{Quick VPC for MCP}}:::announced
B((Enterprise Users)):::external
C([VPC Connection]):::feature
D([MCP Connector]):::feature
E(Private MCP Servers):::compute
F(Amazon EC2):::compute
G(AWS Fargate):::compute
H(AWS Agentcore):::compute
I([Security Controls]):::feature
J((On-Premises Systems)):::external
B ==>|"natural language"| A
A ==>|"routes traffic"| C
C ==>|"connects to"| D
D -->|"invokes tools"| E
E -->|"runs on"| F
E -->|"runs on"| G
E -->|"runs on"| H
C -.->|"Direct Connect/VPN"| J
A -->|"enforces"| I
classDef announced fill:#ff9900,stroke:#ec7211,color:#fff,font-weight:bold
classDef compute fill:#e3f2fd,stroke:#1565c0,color:#1565c0
classDef storage fill:#e8f5e9,stroke:#2e7d32,color:#2e7d32
classDef feature fill:#fff3e0,stroke:#e65100,color:#e65100
classDef external fill:#f5f5f5,stroke:#616161,color:#616161
What's New
Amazon Quick now supports VPC connectivity for Model Context Protocol (MCP) server connections, enabling enterprise customers to link privately hosted MCP servers to the Quick AI assistant without exposing them to the public internet. Previously, Quick's MCP integration was restricted to third-party servers reachable over the public internet, limiting its utility for organizations with sensitive internal tooling. This update allows Quick to securely communicate with MCP servers running on private AWS compute—such as EC2, Fargate, or AWS Agentcore—by routing all traffic through a customer-managed VPC.
How It Works
- VPC Connection Registration: Administrators register a VPC connection within Amazon Quick (Enterprise Edition), which provisions elastic network interfaces inside the customer's VPC, enabling Quick to exchange traffic with private network resources.
- MCP Connector Creation: During MCP connector setup in Quick, the administrator selects an existing VPC connection and supplies the private MCP server URL; no public endpoint is required.
- Traffic Routing: All communication between Quick and the private MCP server is routed exclusively through the customer's VPC, subject to the customer's own route tables, security groups, network ACLs, and subnets.
- Broad Compute Support: MCP servers can run on Amazon EC2, AWS Fargate, AWS Agentcore, or any other compute reachable within the VPC, including on-premises systems connected via AWS Direct Connect, VPN, or VPC peering/Transit Gateway.
- Tool Registration: Each tool exposed by the MCP server is registered as an invocable action in Quick, allowing the AI assistant to call those tools during natural-language conversations on behalf of users.
- Authentication: The integration uses PKCE (Proof Key for Code Exchange) with the S256 challenge method and Resource Indicators (RFC 8707) to bind access tokens to specific MCP servers; OAuth endpoints must remain publicly reachable even when the MCP server itself is private.
- Security Controls: Standard VPC security primitives—CloudTrail logging, security groups, NACLs—apply to all Quick-initiated traffic, giving security teams full visibility and control.
Why It's Important
- Unlocks Private Enterprise Data: Organizations can now expose proprietary databases, internal APIs, and custom tooling to Quick's AI workflows without moving data to public infrastructure, dramatically expanding the assistant's actionable context.
- Eliminates a Critical Security Barrier: The previous requirement for publicly accessible MCP servers was a non-starter for regulated industries (finance, healthcare, government) and security-conscious enterprises; VPC support removes that blocker.
- Preserves Existing Network Controls: Because Quick traffic flows through the customer's VPC, existing perimeter security, compliance logging, and network segmentation policies apply automatically—no new security architecture is needed.
- Enables Hybrid Connectivity: Support for Direct Connect and VPN means on-premises MCP servers (e.g., legacy ERP systems, on-prem databases) can also be integrated, bridging cloud AI capabilities with existing data center investments.
- Accelerates Enterprise AI Adoption: Teams can now connect Quick to internal tools via natural language without building custom integrations, reducing the engineering effort required to operationalize AI across business functions.
How It's Different
- Private vs. Public-Only MCP: Before this update, Quick's MCP support was limited to publicly internet-accessible servers; VPC connectivity is the first path to fully private MCP server integration within Quick.
- Native VPC Integration vs. Proxy Workarounds: Rather than requiring customers to stand up reverse proxies or API gateways to expose private servers publicly, Quick natively integrates with VPC networking primitives.
- Customer-Controlled Network Security: Unlike SaaS-managed tunneling solutions, all traffic stays within the customer's VPC and is governed by the customer's own security groups, NACLs, and routing—not a third-party overlay network.
- Multi-Compute Flexibility: Quick's VPC MCP support is compute-agnostic (EC2, Fargate, Agentcore, on-premises), whereas many competing AI assistant integrations are tightly coupled to specific hosting environments.
- Integrated with Existing Quick Connector UX: VPC MCP setup reuses the existing Quick connector creation workflow, meaning no separate tooling or SDK is required—administrators select a VPC connection and provide a URL.
When to Prefer It
- Regulated Industry Deployments: Use VPC connectivity when operating in healthcare, financial services, or government sectors where data residency and network isolation requirements prohibit exposing internal systems to the public internet.
- Proprietary Internal Tools: Prefer this when your MCP server fronts custom-built internal applications—ticketing systems, ERP modules, proprietary analytics engines—that must never be publicly addressable.
- On-Premises Data Sources: Choose VPC MCP when your data or tooling lives in an on-premises data center connected to AWS via Direct Connect or VPN, and you want Quick to query or act on that data in real time.
- Multi-VPC or Segmented Architectures: Use this when your MCP server resides in a spoke VPC connected via Transit Gateway or VPC peering, and you need Quick to traverse that topology securely.
- High-Sensitivity Custom Data Sources: Prefer VPC connectivity whenever the MCP server provides access to sensitive data (PII, financial records, intellectual property) that requires end-to-end network-level isolation rather than relying solely on application-layer authentication.
- Enterprise Teams Already Using Quick: Ideal for organizations already subscribed to Quick Enterprise that want to extend AI-assisted workflows to internal systems without re-architecting their security posture.
Availability
- GA Status: Generally available as of June 1, 2026; no preview or beta designation indicated.
- Regional Availability: Available in all AWS Regions where Amazon Quick is currently offered; refer to the AWS Regional Services List for the specific region roster.
- Edition Requirement: Requires an Amazon Quick Enterprise subscription; not available on free-tier or non-enterprise plans.
- Pricing: No separate charge announced for VPC connectivity itself; standard Amazon Quick Enterprise pricing applies, with potential VPC data transfer costs governed by standard AWS networking pricing.
- Known Limitation – OAuth Endpoints: Even when the MCP server is private, OAuth endpoints (authorization and token endpoints) used for authentication must remain publicly internet-accessible; private OAuth providers are not supported.
- Known Limitation – Knowledge Bases: Amazon Quick knowledge bases do not currently support VPC integrations; VPC connectivity applies specifically to MCP server connections.
