AWS Security Hub MCP App brings exposure findings into your AI-assisted workflow (Preview)
Investigate AWS Security Hub findings via natural language and interactive attack-path graphs—directly inside Claude Desktop, at no extra cost.
View original announcement →Visual Summary
graph TD
A{{Security Hub MCP App}}:::announced
B((Security Analyst)):::external
C(Claude Desktop):::compute
D(AWS Security Hub):::compute
E([Exposure Findings]):::feature
F([Attack Path Graph]):::feature
G([Network Path View]):::feature
H([Remediation Guidance]):::feature
I(AWS Credentials):::storage
B ==>|"natural language"| C
C ==>|"tool calls"| A
A -->|"read-only APIs"| D
A -->|"authenticates via"| I
D -->|"returns"| E
A -->|"visualizes"| F
A -->|"visualizes"| G
A -.->|"suggests"| H
C -->|"dual response"| B
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
AWS has launched a public preview of the Security Hub MCP App, a local Model Context Protocol (MCP) server that integrates AWS Security Hub exposure findings directly into Claude Desktop. Security engineers can now investigate their security posture using natural language queries—exploring attack paths, network paths, correlated findings, and remediation guidance—without leaving their AI-assisted workflow. The feature is read-only, runs locally using existing AWS credentials, and is available at no additional cost to Security Hub customers.
How It Works
- Local MCP Server: The MCP App runs as a local server on the user's machine, acting as a secure bridge between Claude Desktop and the Security Hub APIs. It uses the machine's existing AWS credential chain and holds no credentials of its own.
- Natural Language Queries: Users interact with Security Hub data through conversational prompts in Claude Desktop, asking questions like "show me my top exposure findings" or "explain the attack path for this finding."
- Dual Response Model: Every tool call returns two outputs simultaneously—a text summary that the AI agent uses for reasoning and an interactive visualization (table, graph, or path diagram) rendered inside Claude Desktop for human verification.
- Available Tools: The MCP App exposes discrete tools covering: top exposures listing, finding detail view (with correlated-finding trait summary), attack path graph (nodes = resources/identities/services, edges = relationships), network path visualization (ordered hops from AWS edge to target resource), and remediation recommendations.
- Exposure Findings: Findings are composite signals that correlate vulnerabilities, misconfigurations, network reachability, and sensitive data exposure into a unified risk view for a given resource.
- Read-Only by Design: Every tool is strictly read-only; no changes are made to the AWS environment or security posture during any investigation session.
Why It's Important
- Reduces Context Switching: Security analysts no longer need to toggle between the Security Hub console, documentation, and separate investigation tools—everything surfaces inside a single Claude Desktop conversation.
- Accelerates Triage: Natural language interaction eliminates the need to manually navigate complex console UIs or write API queries, compressing the time from alert to understanding.
- Human-in-the-Loop Verification: The interactive visualization alongside the AI text summary ensures analysts can visually confirm what the agent is reasoning about, reducing the risk of acting on misinterpreted findings.
- Lowers Barrier to Security Expertise: Junior analysts or developers can ask plain-English questions about attack paths and remediations that would otherwise require deep Security Hub expertise to navigate.
- No Additional Cost: Existing Security Hub customers gain this agentic capability without incremental licensing fees, lowering the adoption barrier for AI-assisted security operations.
- Signals AWS's Agentic Security Direction: This release demonstrates AWS's broader strategy of embedding MCP-based agentic interfaces into core security services, previewing how future security tooling may be consumed.
How It's Different
- MCP-Native Integration vs. Console UI: Unlike the standard Security Hub web console, the MCP App is purpose-built for AI agent consumption, returning structured dual responses (text + visualization) rather than static HTML pages.
- Local Execution vs. Cloud-Side AI: The MCP server runs on the user's machine rather than as a cloud-hosted agent, keeping credential handling local and avoiding data egress concerns associated with cloud-based AI intermediaries.
- Exposure-Centric vs. Finding-Centric: The app focuses on exposure findings—composite, correlated risk views—rather than raw individual findings, providing higher-signal context that maps directly to exploitable risk.
- Interactive Visualizations Inside Chat: Competitors and prior integrations typically return text-only responses; the Security Hub MCP App renders interactive attack path graphs and network path diagrams directly within the Claude Desktop conversation window.
- Read-Only Safety Guarantee: Unlike some agentic security tools that can trigger automated remediations, this app is explicitly read-only, making it safe to deploy in production environments without risk of unintended changes.
- Zero Additional Cost: Many AI-augmented security tools carry premium pricing; this capability is bundled into the existing Security Hub subscription at no extra charge.
When to Prefer It
- Daily Security Triage: Use it when a security analyst starts their day and wants a prioritized, natural-language briefing of the most critical exposure findings across the AWS environment.
- Incident Investigation: Use it when investigating a specific alert and needing to quickly understand the full attack path, network reachability, and correlated signals without navigating multiple console screens.
- Developer Self-Service Security Reviews: Use it when developers want to understand the security posture of resources they own without requiring a dedicated security engineer to interpret Security Hub data for them.
- Security Posture Briefings: Use it when preparing a quick summary of top exposures for a team meeting or stakeholder update, leveraging the AI agent to synthesize findings into plain-language explanations.
- Onboarding New Security Team Members: Use it when ramping up analysts who are unfamiliar with Security Hub's data model, allowing them to ask exploratory questions and learn through interactive visualizations.
- Read-Only Audit Scenarios: Use it in environments where strict change-control policies prohibit any automated remediation tooling, since the read-only guarantee makes it safe for regulated or production contexts.
Availability
- Status: Public preview as of July 27, 2026; not yet generally available (GA).
- Supported Regions: Available in all AWS commercial Regions where AWS Security Hub is supported; GovCloud and China Regions are not included.
- AI Client Requirement: Currently integrates exclusively with Claude Desktop; other MCP-compatible clients are not mentioned as supported at this time.
- Pricing: No additional cost for existing AWS Security Hub customers; standard Security Hub pricing applies.
- Execution Environment: Runs locally on the user's machine; requires a configured AWS credential chain with appropriate Security Hub read permissions.
- Limitations: All tools are read-only—no remediation actions can be executed through the MCP App. Preview status means features and APIs may change before GA.
