IDE Integration

AegisGate Rampart provides real-time PII, secrets, XSS, compliance, and adversarial prompt detection inside your IDE. Three integration paths cover every major editor:

All three connect to a local Rampart proxy running on your machine. Zero external communication. Zero data retention. localhost only.

Prerequisites

Install and start the Rampart proxy:

# Build or download Rampart
./rampart --daemon

# Verify it's running
curl http://localhost:8080/health

Rampart listens on localhost:8080 by default. All IDE plugins connect to this local endpoint.

VS Code / Cursor

Install

1. Open VS Code or Cursor
2. Go to Extensions (Ctrl+Shift+X / Cmd+Shift+X)
3. Search for “AegisGate Rampart”
4. Click Install

Or install from CLI:

code --install-extension aegisgate-rampart
# For Cursor:
cursor --install-extension aegisgate-rampart

Features

Configuration

Open Settings (Ctrl+, / Cmd+,) and search for “aegisgate”:

Commands

Detection Categories

The extension displays 154 regex patterns + ML neural network detections from Rampart:

JetBrains (IntelliJ, PyCharm, WebStorm, GoLand, etc.)

Install

Option A: JetBrains Marketplace (coming soon)

1. Go to Settings → Plugins → Marketplace
2. Search for “AegisGate Rampart”
3. Click Install

Option B: Build from source

git clone https://github.com/aegisgatesecurity/aegisgate-rampart-jetbrains
cd aegisgate-rampart-jetbrains
./gradlew buildPlugin
# Install the .zip from build/distributions/ via:
# Settings → Plugins → ⚙️ → Install Plugin from Disk

Configuration

Go to Settings → Tools → AegisGate Rampart:

Features

• Inline annotations (same icon set as VS Code extension)
• Scan Current File (Tools → Rampart → Scan Current File)
• Check Connection (Tools → Rampart → Check Connection)
• Status bar widget showing detection count + block mode
• Auto-scan on save

Any LSP-Capable Editor (Neovim, Emacs, Helix, Sublime)

The rampart-lsp binary is a Language Server Protocol implementation that communicates over stdio via JSON-RPC 2.0. It calls Rampart’s /detect endpoint and publishes diagnostics back to the editor.

Build

cd aegisgate-rampart
go build -o rampart-lsp ./cmd/rampart-lsp

Usage

./rampart-lsp \
  --rampart-url http://localhost:8080 \
  --debounce-ms 300 \
  --min-severity medium

Neovim Configuration

Add to your init.lua:

-- Configure rampart-lsp as a language server
local lspconfig = require('lspconfig')

lspconfig.rampart_lsp.setup {
  cmd = {
    "rampart-lsp",
    "--rampart-url", "http://localhost:8080",
    "--debounce-ms", "300",
    "--min-severity", "medium"
  },
  filetypes = { "markdown", "text", "plaintext", "python", "javascript", "typescript", "go", "rust", "java" },
  root_dir = function() return vim.fn.getcwd() end,
}

Or in VimScript (init.vim):

" Configure rampart-lsp
lua << EOF
require('lspconfig').rampart_lsp.setup {
  cmd = { "rampart-lsp", "--rampart-url", "http://localhost:8080" },
  filetypes = { "markdown", "text", "python", "javascript", "typescript", "go" },
}
EOF

Emacs Configuration

Add to your init.el:

(use-package lsp-mode
  :hook ((markdown-mode text-mode python-mode) . lsp)
  :config
  (lsp-register-client
   (make-lsp-client
    :new-connection (lsp-stdio-connection
                     '("rampart-lsp" "--rampart-url" "http://localhost:8080"))
    :major-modes '(markdown-mode text-mode python-mode)
    :server-id 'rampart-lsp)))

Helix Configuration

Add to .config/helix/languages.toml:

[language-server.rampart]
command = "rampart-lsp"
args = ["--rampart-url", "http://localhost:8080", "--min-severity", "medium"]

[[language]]
name = "markdown"
language-servers = ["rampart"]

Sublime Text (via LSP package)

Add to LSP.sublime-settings:

{
  "clients": {
    "rampart-lsp": {
      "command": ["rampart-lsp", "--rampart-url", "http://localhost:8080"],
      "enabled": true,
      "languages": [
        { "syntaxes": ["Markdown", "Plain Text", "Python"] }
      ]
    }
  }
}

How It Works

You type in IDE
     │
     ▼
IDE plugin (VS Code / JetBrains / LSP)
     │
     ▼ HTTP POST to localhost
Rampart proxy (running locally as --daemon)
     │
     ▼ 154 regex patterns + CharCNN-BiLSTM ML model
Detection results
     │
     ▼ JSON-RPC diagnostics
Back to IDE as inline warnings

1. You type or paste text in your editor
2. The plugin sends text to Rampart’s /detect endpoint on localhost
3. Rampart runs 154 regex patterns + ML neural network detection
4. Results return as LSP diagnostics (inline warnings with severity + category)
5. You see 🔐 🔑 ⚔️ 📋 🧠 icons next to detected content

No text ever leaves your machine. The plugin talks only to localhost. Rampart does the detection locally. No cloud, no telemetry, no external API calls.

Privacy (12 Non-Negotiables)

All three IDE integrations enforce the same privacy rules:

1. No prompt text stored or sent externally
2. No URLs logged
3. No page content stored
4. No PII stored
5. No credentials stored
6. No fingerprinting
7. No cross-site tracking
8. No provider metadata collected
9. No keystroke logging
10. No clipboard reading
11. No file system access beyond what the editor provides
12. All communication is localhost-only

Troubleshooting

“No detections appearing”

1. Verify Rampart is running: curl http://localhost:8080/health
2. Check the proxy URL in your IDE settings matches Rampart’s port
3. Try a manual scan (Command Palette → “Rampart: Scan Active Document”)
4. Test with known PII: paste My SSN is 123-45-6789 — should trigger a PII detection

“Connection failed”

1. Check Rampart is running with --daemon flag
2. Verify no firewall is blocking localhost connections
3. Try curl http://localhost:8080/detect -d '{"text":"test"}'

“JetBrains plugin: wrong port”

The JetBrains plugin defaults to port 9090. If Rampart runs on 8080 (default), update Settings → Tools → AegisGate Rampart → Rampart proxy URL to http://localhost:8080.

“VS Code: too many diagnostics”

Lower aegisgate-rampart.maxDiagnostics or raise aegisgate-rampart.severityThreshold to medium or high.