MCP-native security proxy · v0.1

The security layer
MCP forgot

Vedis sits between your AI agents and MCP servers. Detects prompt injection, enforces tool policies, filters secrets. One config change, zero code changes.

$ npm install -g vedis click to copy
View on GitHub
73%
of MCP tool calls are vulnerable to prompt injection attacks
43%
of MCP servers ship with auth or command injection flaws
0
existing MCP-native security proxies — until now
Architecture
How it works
Vedis is a transparent proxy. No SDK, no code changes. Point your agent at Vedis, point Vedis at your MCP server.

Agent

Sends MCP tool call
via stdio or SSE

🛡

Vedis

Scans, filters, enforces

  • Injection scanner
  • Policy engine
  • Output filter

MCP Server

Receives clean,
policy-compliant request

Capabilities
Everything you need to lock down MCP
Five middleware modules. All configurable. All optional. Mix and match.
🔎

Prompt Injection Scanner

20+ heuristic patterns for override attempts, role hijacking, encoding tricks. Compound threat scoring with 3 sensitivity levels — low, medium, high.

📜

Tool Policy Engine

YAML-based allowlist and blocklist rules. Glob patterns for tool matching. Fine-grained constraints on arguments, methods, and resources.

🚫

Output Filter

PII detection: emails, phone numbers, SSNs, credit cards. Secret scanning: AWS keys, GitHub tokens, Stripe keys, JWTs, PEM certificates.

📑

Audit Logger

JSONL and SQLite backends. Indexed queries over full request/response trails. Every decision logged with timestamps and threat details.

Rate Limiter

Sliding window algorithm with configurable per-minute limits. Returns proper JSON-RPC error responses. Prevents runaway agent loops.

📈

Live Dashboard

Real-time stats and metrics. Interactive injection scanner test. Audit log viewer with filtering. All served from a built-in HTTP server.

Get started
Three commands. That's it.
Install globally, generate a config, start the proxy. Your agent connects to Vedis instead of the MCP server directly.
  terminal
# Install Vedis globally
$ npm install -g vedis

# Generate default config
$ vedis init

# Start the security proxy
$ vedis proxy --upstream "npx -y @modelcontextprotocol/server-filesystem /tmp"
Pricing
Start free. Scale when you need to.
The open source core is fully featured. Pro and Team add cloud-hosted dashboards and advanced detection.
Open Source
Free
Everything you need to secure a single agent.
  • Core proxy (stdio + SSE)
  • All 5 middleware modules
  • JSONL audit logging
  • Built-in dashboard
  • Community support
Get Started
Team
$199/mo
For organizations running multiple agents at scale.
  • Everything in Pro
  • 1M requests/month
  • SSO integration
  • Team dashboard
  • Slack / Discord alerts
  • Priority support