Infrawise MCP Server

Community

Give your AI coding assistant deterministic awareness of your AWS and database infrastructure

How to install ↓ Compare GitHub source repository ↗

Category: Dev Tools #60 of 438
Stars: ★ 22 Popular
Transport: stdio (local process) · Streamable HTTP
Runtime: Node.js 22+ · Python 3+
Credentials: API key / credential required
License: MIT
Last commit: 7 days ago
Tools: 21

64FMRS · C

A well-designed, read-only, local-first deterministic infrastructure analysis MCP server with 21 tools covering AWS, three databases, Kafka, and IaC drift — ideal for grounding AI coding assistants. Code-level correlation is currently limited to TS/JS/Python.

Strongest · Documentation 17/20 Weakest · Reliability 8/20

Reliability

8/20

Security and permissions

12/20

Maintenance

14/20

Documentation

17/20

Setup experience

13/20

Why each score

Reliability 8/20: Only CI/npm-publish badges are referenced in the README; no actual workflow content or committed tests were provided to verify that the 21 declared tools match real behavior, or that the MCP init/tool-list handshake is reproducible. Per the static calibration rule, reliability cannot exceed 12 without verifiable execution evidence, so 8 is given — the deduction reflects missing evidence, not an assumed best case.
Security and permissions 12/20: The README claims read-only operation, least privilege, credentials never stored, and no telemetry, and backs this with a minimal DynamoDB IAM policy example and ${ENV_VAR} substitution to avoid committing plaintext passwords — good practice if true. However, these are self-reported claims not verifiable against source code in this review, and there is no described confirmation step for potentially risky misconfiguration (e.g., accidentally pointing at a production database). Auto-executing a third-party package via `npx -y` is standard MCP practice, not a red line. Overall the main risks are disclosed but not independently verifiable, so full marks are not warranted.
Maintenance 14/20: The repo is not archived, has a clear MIT license, and version 0.28.1 suggests multiple iterative releases; README documents a full CONTRIBUTING workflow and release process (pnpm release) plus a linked GitHub Project roadmap. However, stars (19, not scored) and only 1 open issue provide limited signal, no explicit security-response channel is mentioned, and maintainer/team scale is unclear — active but with governance gaps.
Documentation 17/20: Documentation is extensive: installation, quick start, a full 21-tool reference table, CLI reference, complete config example, AWS least-privilege policy, a PostgreSQL read-only user setup script, categorized analysis capabilities, an explicit current-limitations section, a security section, and source layout are all covered, with limitations disclosed (e.g., Python requires python3, dynamic queries may not resolve statically). Points are deducted for lacking a dedicated troubleshooting section and cross-version compatibility notes, and for no reviewable test evidence beyond narrative screenshots/gifs.
Setup experience 13/20: The install path is clearly described: `npm install -g infrawise` or `npx infrawise start --claude`, which auto-writes `.mcp.json`/`.cursor/mcp.json`/`.vscode/mcp.json` with example configs for Claude Code, Cursor, and VS Code — a concise flow. Per the static calibration rule, setup cannot exceed 15 without verifiable execution evidence (real run logs, test coverage); additionally, first-run environment probing (AWS/DB discovery, IAM permissions, connection strings) likely adds real-world friction beyond the documented one-liner, so 13 is given.

Static review · not runListed 2026-09-07

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesUses the networkConnects to a database

Best for

Teams building on AWS with TypeScript/JavaScript/Python
Developers who want AI assistants to ground code generation in real infrastructure
Engineering teams wanting infrastructure anti-pattern gates in CI

Not for

Projects not using AWS or the supported databases/messaging services
Tooling that must provision or modify infrastructure (Infrawise is read-only)
Code-level correlation in languages other than TypeScript/JS/Python (infrastructure-level analysis still works)

Required permissions

AWS credentials via your existing credential chain (minimum dynamodb:ListTables/DescribeTable; full read-only IAM policy in docs)
Optional read-only PostgreSQL/MySQL/MongoDB connections (read-only user recommended)
SSO profiles require aws sso login first

Risks and side effects

Reads infrastructure metadata via AWS APIs and DB connections; read-only and local, but use least-privilege credentials
Python code scanning requires python3 on PATH
Some supply-chain scanners may flag the 'infra' prefix as deceptive naming (author states this is a false positive; project is independent and unaffiliated)
Large monorepos may need future incremental analysis optimization

Setup

Before you start

Runtime：Node.js 22+ · Python 3+

DB_PASSWORD optionalsecret Database password substituted into the postgres connectionString via ${DB_PASSWORD} in infrawise.yaml, so credentials are never committed in plaintext.

Requires Node.js 22+. Run npm install -g infrawise or use npx infrawise start --claude (also --cursor, --vscode). The start command generates infrawise.yaml, scans your infrastructure, and writes the editor MCP config. The editor then manages infrawise serve --stdio; for HTTP run infrawise serve and configure url http://localhost:3000/mcp.

.mcp.json

{"mcpServers":{"infrawise":{"command":"npx","args":["-y","infrawise","serve","--stdio"]}}}

Shown for Claude Code. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json

{
  "servers": {
    "infrawise": {
      "command": "npx",
      "args": [
        "-y",
        "infrawise",
        "serve",
        "--stdio"
      ]
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal

claude mcp add infrawise -- npx -y infrawise serve --stdio

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

Run infrawise start --claude and launch Claude Code; the tool list should show the 21 tools such as get_infra_overview, and calling get_infra_overview should return a services snapshot with a freshness flag.

Troubleshooting

Run infrawise doctor to validate AWS/DB access, config, and repo scan
Every service must be explicitly enabled: true in config; unlisted services are never contacted
Cache expires after 24 hours; run infrawise analyze to force a rescan after major changes
Dynamically constructed queries may not resolve statically
Re-probe with infrawise start --rediscover

Things to try

Once connected, you can ask your AI assistant things like:

What partition keys, sort keys, and GSIs does my Orders DynamoDB table have?
Analyze the listAllOrders function for full table scans or missing indexes.
Give me the exact CREATE INDEX SQL for the missing email index on the users table.
Which SQS queues lack DLQs or are unencrypted?

Tools 21

get_infra_overview read-only

Full snapshot — services, counts, high-severity findings, analysis freshness, and configured flag

get_graph_summary read-only

Full infrastructure graph — all nodes, edges, and findings

get_table_schema read-only

Column-level schema for named tables/collections — types, PKs, FKs, indexes, DynamoDB keys/billing mode, cost signal

analyze_function read-only

Issues in a specific function — scans, missing indexes, N+1, trigger event shapes, missing IAM permissions

suggest_gsi read-only

Exact GSI config for a DynamoDB table + attribute

postgres_index_suggestions read-only

Exact CREATE INDEX SQL for your actual PostgreSQL table

suggest_mongo_index read-only

Exact createIndex command for a MongoDB collection + field

mysql_index_suggestions read-only

Exact ALTER TABLE ADD INDEX SQL for your MySQL table

Show 13 more tools

get_queue_details read-only

SQS queues — DLQ status, encryption, FIFO type, visibility timeout, message counts

get_api_routes read-only

API Gateway APIs (REST/HTTP/WebSocket) — routes, methods, paths, and Lambda integrations

get_topic_details read-only

SNS topics — subscription counts, protocols, and filter policies

get_secrets_overview read-only

Secrets Manager — names, rotation status, and inferred key names (values never included)

get_parameter_overview read-only

SSM Parameter Store — names, types, tiers (values never included)

get_lambda_overview read-only

Lambda functions — runtime, memory, timeout, role ARN, triggers, env var key names, cost signal

get_eventbridge_details read-only

EventBridge rules — name, state, schedule/event pattern, target functions

get_s3_overview read-only

S3 buckets — versioning, encryption, public access, event notifications

get_log_errors read-only

CloudWatch error patterns and counts (no raw log messages)

get_stack_outputs read-only

Stack outputs and cross-stack exports parsed from local IaC files

get_cognito_overview read-only

Cognito user pools — MFA config, auth flows, OAuth settings, token validity

get_stream_details read-only

Kinesis streams (shards, retention, capacity mode) and MSK clusters (state, Kafka version, brokers)

get_cache_overview read-only

ElastiCache clusters — engine, encryption, replication, failover, cost signal

Use cases

Get real table/index context in Claude Code before generating DynamoDB or PostgreSQL queries

Detect full scans, N+1, hot partitions, and other anti-patterns in code

Gate deployments in CI with infrawise check severity thresholds

Detect drift between IaC files and deployed state

Supported clients

Claude Code

Cursor

GitHub Copilot

Listed from the project's documentation, not tested by this site.

Overview

Infrawise is an MCP server for infrastructure analysis. It statically analyzes your codebase, AWS resources, database schemas, and IaC files (Terraform/CloudFormation/CDK), uses a graph engine and 36 rule-based analyzers to detect anti-patterns like full table scans, missing indexes/GSIs, hot partitions, N+1 queries, and missing DLQs, and exposes that context through 21 MCP tools to editors such as Claude Code, Cursor, and GitHub Copilot. Analysis is fully local, read-only, deterministic (no LLM involved in analysis), and telemetry-free. Supports stdio and HTTP (Streamable HTTP on port 3000) transports.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 63.3k · Tools 2 Compare with this →

NPM Sentinel MCP Server 79 · B

AI-powered NPM package analysis MCP server

★ 18 · Tools 19 Compare with this →

MartinLoop MCP Server 79 · B

An independent control layer that gives AI coding agents budgets, brakes, and receipts

★ 413 Compare with this →

GitHub MCP Server 76 · B

GitHub's official MCP server for managing repos, issues, PRs, and workflows via natural language

★ 33.7k · Tools 15 Compare with this →

Compare with other MCP servers

Source revision 3b590389a92a Data synced 2026-10-19 Read the FMRS scoring method