AegisGate Rampart — Local AI Security Proxy & IDE Plugin

Free, open-source local proxy that intercepts AI traffic from Copilot, Cursor, and any LLM. Real-time detection in VS Code, JetBrains, Neovim, and any LSP editor. Scans prompts and responses for secrets, PII, and injection attacks before they reach the model.

⚡ AegisGate Rampart v0.7.1 is LIVE — Local proxy, IDE plugins for VS Code/Cursor and JetBrains, real-time detection. Download from GitHub (free, open source, Apache 2.0).

⚡ AegisGate Rampart is free and open source. No account required. No cloud calls. All detection happens locally on your machine.

Download Rampart v0.7.1 → Quick start guide →

What is Rampart?

Rampart is a local security tool for developers who use AI coding assistants like GitHub Copilot, Cursor, or local LLMs (Ollama, LM Studio). It sits between your editor and the AI model, scanning everything you send — and everything the AI sends back — for sensitive data and security risks.

Think of it as a firewall for AI coding tools. It catches the moment you’re about to send a database password to Copilot, or when the AI generates code that contains an API key, and stops it before it’s too late.


Do I need Rampart?

If you use AI chat in a browser (ChatGPT, Claude, etc.): You need Lens — it’s a free browser extension that protects you in the browser. Rampart is not for this use case.

If you use AI coding tools (Copilot, Cursor, local LLMs, API calls): You need Rampart. It protects the places Lens can’t reach — your IDE, your terminal, your API calls.

Your setupWhat to use
ChatGPT or Claude in a browserLens (free browser extension)
GitHub Copilot in VS CodeRampart (IDE plugin)
Cursor AI editorRampart (IDE plugin)
Local LLMs (Ollama, LM Studio)Rampart (local proxy)
API calls to OpenAI/Anthropic from your codeRampart (local proxy)
Both browser chat AND coding toolsLens + Rampart (both are free)

See Rampart in Action

Watch Rampart intercept AI traffic from VS Code, detect hardcoded secrets, and protect your coding workflow.


What does it catch?

RiskExamples
SecretsAPI keys (AWS, GitHub, OpenAI, Stripe), database passwords, SSH private keys, JWT tokens, OAuth tokens
Personal info (PII)SSN, email, phone, credit card, passport, bank routing numbers
Prompt injectionAdversarial prompts designed to make the AI ignore safety rules, leak system prompts, or execute unauthorized actions
Compliance violationsText that violates HIPAA, GDPR, PCI-DSS, EU AI Act
Malicious code (XSS)Script injection, event handlers, encoded payloads, SVG vectors
Response risksPII leaked in AI responses, hallucinated secrets, injected content in model output

Architecture Overview

flowchart TD subgraph IDE["IDE Integration"] VS[VS Code / Cursor] JB[JetBrains] NV[Neovim / LSP] end subgraph Rampart["AegisGate Rampart v0.6.2"] direction TB subgraph PluginMode["IDE Plugin Mode"] PM1[LSP Server] PM2[Real-time Detection] PM3[Inline Warnings] end subgraph ProxyMode["Local Proxy Mode"] PX1[MITM Proxy :8443] PX2[Traffic Interception] PX3[API Redirection] end subgraph Engine["Detection Engine"] E1[176+ Patterns] E2[PII + Secrets] E3[XSS + Compliance] E4[ML Adversarial] end end subgraph AI["AI Services"] A[Copilot] B[Cursor] C[Ollama] D[OpenAI API] E[Local LLMs] end VS --> PluginMode JB --> PluginMode NV --> PluginMode PluginMode --> Engine ProxyMode --> Engine PluginMode --> AI ProxyMode --> A ProxyMode --> B ProxyMode --> C ProxyMode --> D ProxyMode --> E style PluginMode fill:#1a1f2e,stroke:#f59e0b,stroke-width:2px style ProxyMode fill:#1a1f2e,stroke:#f59e0b,stroke-width:2px style Engine fill:#1a1f2e,stroke:#38bdf8,stroke-width:2px

Detection Flow

sequenceDiagram participant Dev as Developer participant IDE as IDE / Editor participant Rampart as Rampart participant Proxy as Local Proxy participant AI as AI Service Dev->>IDE: Write Code / Prompt IDE->>Rampart: Send for Analysis (LSP) Rampart->>Rampart: Scan (PII, Secrets, XSS) alt Threat Detected Rampart-->>IDE: Show Warning (Inline) Dev->>IDE: Edit & Remove Risk IDE->>Rampart: Re-scan else Clean Rampart-->>IDE: Allow Send IDE->>Proxy: Forward to AI Proxy->>AI: POST Request AI-->>Proxy: AI Response Proxy->>Proxy: Scan Response Proxy-->>IDE: Return Response end Note over Rampart: All detection local (~5ms) Note over Proxy: Zero data leaves machine

Is it private?

Yes. Everything runs on your machine. Nothing is sent to any server — not your prompts, not your code, not your detection results.

✅ Rampart DOES❌ Rampart does NOT
Run entirely on your machineSend your prompts or code to any server
Scan traffic locally in real-timePhone home with analytics or telemetry
Encrypt its CA keys with a passphraseStore plaintext keys on disk
Redact PII/secrets from audit logsLog your actual prompt text or secrets
Stay free and open sourceRequire an account or subscription

Open source (Apache 2.0): github.com/aegisgatesecurity/aegisgate-rampart



Quick Start

VS Code / Cursor:

  1. Open the Extensions panel (Ctrl+Shift+X / Cmd+Shift+X)
  2. Search for “AegisGate Rampart”
  3. Click Install, then Reload
  4. Detection runs automatically as you type — results appear as inline warnings

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

  1. Open Settings → Plugins → Marketplace
  2. Search for “AegisGate Rampart”
  3. Click Install, then Restart
  4. Detection results appear in the Problems tool window

Neovim / any LSP editor:

# Download the Rampart LSP binary
curl -L https://github.com/aegisgatesecurity/aegisgate-rampart/releases/latest/download/rampart-lsp-linux-amd64 -o /usr/local/bin/rampart-lsp
chmod +x /usr/local/bin/rampart-lsp

# Add to your LSP config (Neovim example)
# lspconfig.rampart_lsp.setup({})

Option 2: Run as a local proxy

# Download for your platform
curl -L https://github.com/aegisgatesecurity/aegisgate-rampart/releases/latest/download/aegisgate-rampart-linux-amd64 -o aegisgate-rampart
chmod +x aegisgate-rampart

# Start the proxy
./aegisgate-rampart --port=8443 --upstream=https://api.openai.com

# Point your AI tool at the proxy
# Instead of: https://api.openai.com/v1/chat/completions
# Use:        http://localhost:8443/v1/chat/completions

Option 3: Docker

docker run -d \
  -p 8443:8443 \
  -p 9090:9090 \
  ghcr.io/aegisgatesecurity/aegisgate-rampart:v0.7.1

How it works

Rampart runs in two modes — you can use either or both:

1. IDE Plugin Mode

Rampart runs inside your editor as a Language Server (LSP). As you type prompts or code that will be sent to an AI tool, Rampart checks for sensitive data and security risks in real-time. Warnings appear inline, just like lint errors.

EditorPluginStatus
VS Code / Cursoraegisgate-rampart-extv0.3.0, published (IDE plugins — Rampart core is v0.6.2)
JetBrains (IntelliJ, PyCharm, etc.)aegisgate-rampart-jetbrainsv0.3.0, published (IDE plugins — Rampart core is v0.6.2)
Neovim / Emacs / Helix / SublimeRampart-LSP (generic LSP server)Available via rampart-lsp binary

The LSP server uses JSON-RPC 2.0 over stdio — it works with any editor that supports LSP.

2. Local Proxy Mode

Rampart runs as a transparent proxy on your machine. Any AI tool that makes HTTP requests — Copilot, Cursor, local LLMs, API calls — gets intercepted and scanned.

Your AI tool (Copilot/Cursor/API)  Rampart proxy (localhost:8443)  AI model (OpenAI/Anthropic/local)
  • Rampart generates a local CA certificate and configures your system to trust it
  • All HTTPS traffic to AI services flows through the proxy
  • Requests and responses are scanned in real-time
  • Malicious content is blocked before it reaches the model (or before the response reaches you)
  • CA keys are encrypted at rest with a passphrase — no plaintext keys on disk

Rampart vs Lens vs Platform

FeatureLens (browser)Rampart (local proxy/IDE)Platform (server)
Protects browser AI chat
Protects IDE AI tools
Protects API calls
Protects server-to-server AI
Real-time editor detection
Compliance frameworks5 facets5 facets31 frameworks
SIEM/SOAR integration
Multi-tenant
Air-gapped deployment
PriceFreeFreeFree tier + paid


For Organizations

Typical team setups

  • Individual developer: Lens (browser) + Rampart (IDE) → full local protection, zero cost
  • Developer team (2-10): Rampart (IDE) for each dev + Platform (server) for centralized policy, audit logs, and compliance reporting
  • Enterprise: Platform (server) + Lens (browser) for non-technical staff + Rampart (IDE) for developers + SIEM/SOAR integration + 31 compliance frameworks

Privacy & Security (for compliance teams)

  1. Zero data collection — no telemetry, no analytics, no phone-home
  2. All detection is local — patterns and ML model run on your machine
  3. CA keys encrypted at rest — passphrase-protected, never plaintext
  4. Audit log redaction — PII and secrets stripped from logs before writing to disk
  5. No network egress — Rampart never sends your data anywhere
  6. Open source — Apache 2.0, auditable, no hidden behavior

Download

Release Assets (v0.7.1)

PlatformAssetArchitecture
macOSaegisgate-rampart-darwin-amd64.tar.gzIntel
macOSaegisgate-rampart-darwin-arm64.tar.gzApple Silicon
Linuxaegisgate-rampart-linux-amd64.tar.gzx86_64
Linuxaegisgate-rampart-linux-arm64.tar.gzARM64
Linuxaegisgate-rampart-linux-amd64.debDebian/Ubuntu
Linuxaegisgate-rampart-linux-amd64.rpmRHEL/Fedora
Windowsaegisgate-rampart-windows-amd64.zipx86_64
Dockerghcr.io/aegisgatesecurity/aegisgate-rampart:v0.7.1Multi-arch (cosign-signed)

All 13 assets on GitHub → Source code →


IDE Integration Details

For detailed setup instructions for each editor, see our IDE Integration Guide.

VS Code / Cursor setup
  1. Install the AegisGate Rampart extension from the VS Code marketplace
  2. The extension automatically starts the Rampart LSP server
  3. Detection results appear as inline diagnostics (like lint warnings)
  4. Configure severity levels in .vscode/settings.json:
{
  "rampart.detection.severity": "warning",
  "rampart.detection.categories": ["pii", "secrets", "xss", "compliance", "ml"]
}
JetBrains setup
  1. Open Settings → Plugins → Marketplace
  2. Search “AegisGate Rampart” → Install → Restart
  3. Detection results appear in the Problems tool window
  4. Configure in Settings → Tools → AegisGate Rampart
Neovim setup (any LSP editor)
-- Neovim (init.lua with nvim-lspconfig)
require'lspconfig'.rampart_lsp.setup{
  cmd = {"/path/to/rampart-lsp"},
  filetypes = {"*"}, -- scan all file types
  settings = {
    detection = {
      categories = {"pii", "secrets", "xss", "compliance", "ml"},
      severity = "warning"
    }
  }
}

Architecture

graph TB subgraph "Your Machine" A[AI Tool
Copilot/Cursor/API] --> B[Rampart Proxy
localhost:8443] A --> C[IDE Plugin
VS Code/JetBrains/LSP] B --> D[Detection Engine] C --> D D --> E{Block or Allow?} E -->|Block| F[🛑 Warning + Log] E -->|Allow| G[Forward to AI Model] end G --> H[AI Model
OpenAI/Anthropic/Local] H --> I[AI Response] I --> B B --> J[Response Scanner] J --> K{Safe?} K -->|No| F K -->|Yes| L[Return to AI Tool]

Frequently Asked Questions

Does Rampart slow down my AI tools?

No. Detection adds ~5ms latency, which is imperceptible compared to the seconds-long response times of AI models. The ML model runs in pure Go — no external process, no network calls.

Does Rampart store my prompts?

No. Rampart processes prompts in memory and discards them. Audit logs are written with PII and secrets redacted — only the detection category and severity are recorded.

Can I use Rampart without the proxy?

Yes. The IDE plugin mode (LSP) works without the proxy. You can also use just the proxy without the IDE plugin. They’re independent.

How is Rampart different from Lens?

Lens is a browser extension that watches what you type into AI chat websites. Rampart is a system-level proxy and IDE plugin that catches AI traffic from any application — Copilot, Cursor, API calls, local LLMs. They use the same detection engine but protect different surfaces.

Is Rampart free for commercial use?

Yes. Apache 2.0 license. Free for personal and commercial use. No restrictions, no attribution required beyond the license terms.

Canonical facts (v0.7.1)

Source: aegisgate-rampart repo

  • Local proxy: Intercepts AI traffic transparently at the system level — no app changes needed
  • IDE integration: VS Code/Cursor extension, JetBrains plugin, generic LSP server (Neovim, Emacs, Helix, Sublime)
  • MITM block mode: Blocks malicious prompts before they reach the AI model
  • CA key encryption: Encrypted at rest with passphrase — no plaintext keys on disk
  • Audit log redaction: PII and secrets stripped from logs before writing
  • 1,318 test functions, 80.7% coverage
  • 13 release assets: macOS (Intel + ARM), Linux (deb + rpm), Windows, Docker (multi-arch, signed)
  • Apache 2.0, zero external dependencies for core functionality