aws-bleed

5-second read-only scan for expensive AWS hidden taxes: NAT Gateway traffic to S3/DynamoDB without free gateway endpoints, CloudWatch Logs with infinite retention, gp2 and orphaned EBS volumes, idle Elastic IPs, and stale snapshots, complete with USD/month estimates and a fix-bleed.tf Terraform patch.

Install

Bun / npm:

bunx aws-bleed
npx aws-bleed

# or install in your project / globally
bun add -d aws-bleed
npm i -D aws-bleed

JSR:

bunx jsr add @kanchingyg/aws-bleed
npx jsr add @kanchingyg/aws-bleed

Go (native binary):

go install github.com/kanchingyg/aws-bleed@latest

The core engine is Go (main.go, internal/). TypeScript under src/*.ts is included for npx / bunx / JSR usage and MCP bridge helpers when no native binary is present.

Related MCP server: AWS Cost Saver

Quick start

npx aws-bleed
bunx aws-bleed
npx aws-bleed --patch fix-bleed.tf --allow-write
npx aws-bleed --mcp

What you get

MCP + local models

npx aws-bleed --mcp
bunx aws-bleed --mcp

Use summarize_exec_report with --model ollama/llama3.2 (or any OpenAI-compatible local server) so cost data stays on your machine.

Requirements

• Node 18+ or Bun for npx / bunx, or Go 1.22+ for native CLI

• AWS credentials with read-only Describe* / List* (no writes unless you pass --allow-write for patch files)

MCP Tools Reference

Editor and Agent Integration

Cursor (.cursor/mcp.json)

{
  "mcpServers": {
    "aws-bleed": {
      "command": "npx",
      "args": ["-y", "aws-bleed", "--mcp"]
    }
  }
}

Claude Code CLI

claude mcp add aws-bleed -- npx -y aws-bleed --mcp

Programmatic Usage (JSR / TypeScript)

import * as bridge from "jsr:@kanchingyg/aws-bleed/bridge";

Runtime Safety and Guardrails

• Stdio frame isolation: When --mcp is active, hijackStdoutForMcp() redirects accidental console.log calls to stderr so JSON-RPC 2.0 frames on stdout never corrupt.

• Read-only by default: Mutating tools (mutates: true) are blocked in --dry-run mode and require explicit --allow-write authorization.

• Loop breaker: Repeated identical (toolName, args) invocations are hashed with SHA-256 and halted after 3 consecutive repeats to prevent runaway agent loops.

• Context budget truncation: Tool outputs are middle-truncated (60% head / 40% tail) scaled to the probed local model n_ctx (Ollama /api/show, llama.cpp /props, vLLM /v1/models).

• Secret redaction and path normalization: AWS keys (AKIA*, ASIA*), bearer tokens, and sk-* keys are scrubbed before returning tool output, and Windows drive paths (C:\...) translate cleanly to /mnt/c/... when WSL mode is enabled.

License

MIT