Most LLMs lack reliable, up-to-date access to cloud provider documentation and APIs; this repo fills that gap by packaging many AWS-focused MCP servers into an opinionated, installable library. Instead of embedding stale knowledge or unsafe API calls in prompts, MCP servers deliver live docs, code patterns, and validated tool interfaces directly into a model's context so assistants can generate accurate answers and, where allowed, propose or execute AWS operations with traceability.
What Sets It Apart
- Standardized MCP integration: Each server implements the Model Context Protocol so a single LLM client (IDE assistant, chatbot, or agent) can connect to multiple capability-specific servers (docs, IaC, Bedrock KBs, DB access, deployments) using the same client-side plumbing. This reduces bespoke connector work and improves interoperability.
- AWS-first, breadth of capabilities: The collection maps common developer workflows (documentation lookup, IaC guidance, CloudFormation/CDK/Terraform helpers, Bedrock & Kendra retrieval, Lambda tool invocation, EKS/ECS management, pricing & cost tools, etc.) to discrete MCP servers so assistants can pull targeted context for each task.
- Managed vs local options + governance: AWS offers remote/managed MCP endpoints (e.g., AWS MCP Server and AWS Knowledge MCP) for teams that want centrally governed, audited access; the repo also provides local container and stdio-based servers for privacy-sensitive or offline workflows.
- Protocol and transport notes: The servers in this repo are designed for stdio transport; Server-Sent Events (SSE) support was removed in the major 2025 releases (May 26, 2025) and Streamable HTTP support is planned per MCP spec—important when integrating with different clients.
How it works (short)
MCP servers run sidecar-like processes that expose structured capabilities (search docs, call validated AWS APIs, run deploy checks, fetch pricing) over the Model Context Protocol. LLM clients maintain 1:1 connections and request context fragments or invoke tool-like operations; servers reply with context slices, citations, or syntactically validated commands. For sensitive actions (writes, provisioning), servers and clients rely on IAM, optional auto-approval toggles, and CloudTrail/audit logging when using AWS-managed endpoints.
Who it's for—and tradeoffs
Great fit if you: developers or teams want to augment LLM assistants with authoritative AWS docs and actionable tooling (RAG-enabled chat, code generation tied to current APIs, IaC guidance, automated runbook/agent workflows), or if you need centralized governance (managed MCP endpoints) while preserving fine-grained controls. Look elsewhere if you: require a pure HTTP/REST streaming transport today for all clients (many repo servers currently use stdio only), must avoid running any local processes or containers (some servers are local-only), or want an off-the-shelf multi-cloud MCP catalog—this repo targets AWS-centric capabilities and includes several deprecated servers with recommended replacements.
Practical notes & limitations
- Created as a modular catalog: pick only the servers you need (e.g., aws-documentation-mcp-server, aws-iac-mcp-server, bedrock-kb-retrieval-mcp-server). Many servers include example one-click installs for Kiro, Cursor, and VS Code.
- Security & credentials: some servers require AWS credentials, IAM roles, or resource ARNs; managed AWS endpoints can avoid credential exposure and provide CloudTrail auditing, but local servers put responsibility for secrets and least-privilege on the user.
- Lifecycle: several servers are marked deprecated in favor of consolidated replacements (e.g., IaC server instead of older CDK/Cloud Control API servers). Transport capabilities are evolving (SSE removed 2025-05-26; Streamable HTTP is planned), so integration tests are advised.
In short, awslabs/mcp is a practical, AWS-focused MCP server collection designed to inject current documentation, validated APIs, and domain tools into LLM workflows—trading a small ops footprint (local processes or container images) for more accurate, auditable, and actionable AI assistants when working with AWS.
