// Documentation

Interlucid Docs

Everything you need to deploy Interlucid as a security proxy in your AI agent infrastructure. Full documentation ships with beta access.

01 / Overview

Interlucid is a transparent security proxy for MCP (Model Context Protocol) traffic. It sits inline between your AI agents and the MCP servers they call — inspecting every request, enforcing your policies, and logging every action to an immutable audit trail.

No changes are required to your agents or MCP servers. You route traffic through Interlucid via a single configuration change, and security visibility is immediate.

02 / How It Works

Agent makes a tool call

Your AI agent (Claude, GPT-4, custom, etc.) invokes an MCP tool — e.g. reading a file, querying a database, sending an email.

Traffic routes through Interlucid

Instead of going directly to the MCP server, the request passes through your Interlucid proxy endpoint. No code changes needed — just update the server URL in your agent config.

Policy engine evaluates the request

Interlucid inspects the full request context — agent identity, tool name, parameters, and content — and evaluates it against your policy rules in real time.

Allow or block — and log everything

Permitted requests are forwarded to the MCP server. Blocked requests are rejected with a structured error. Every action — allowed or blocked — is logged to your immutable audit trail.

03 / Quickstart

The following is a preview of the Interlucid setup flow. Full configuration details are provided during beta onboarding.

1. Install the Interlucid proxy

BASH

# Install via npm
npm install -g @interlucid/proxy

# Or run via Docker
docker pull interlucid/proxy:latest

2. Create your config file

YAML — interlucid.config.yml

proxy:
  port: 4080
  api_key: "il_your_api_key_here"

upstream:
  url: "http://your-mcp-server:3000"

logging:
  level: "full"   # full | metadata | minimal
  retention_days: 90

policies:
  default_action: "allow"
  rules:
    - name: "block-sensitive-files"
      match:
        tool: "filesystem.read"
        path_pattern: "/etc/*"
      action: "block"
    - name: "block-external-exfil"
      match:
        tool: "network.fetch"
        url_not_in: "allowlist"
      action: "block"

3. Start the proxy

BASH

interlucid start --config interlucid.config.yml

# Output:
# ✓ Interlucid proxy running on http://localhost:4080
# ✓ Policy engine loaded — 2 rules active
# ✓ Audit log streaming to interlucid.ai/dashboard

4. Point your agent at the proxy

BEFORE / AFTER

# Before — agent calls MCP server directly
MCP_SERVER_URL="http://your-mcp-server:3000"

# After — agent calls Interlucid proxy instead
MCP_SERVER_URL="http://localhost:4080"

04 / Core Concepts

🛡 The Proxy

A transparent HTTP/S proxy that intercepts MCP tool calls. Adds sub-5ms overhead. Supports all MCP server implementations.

📋 Audit Log

Immutable, tamper-proof log of every tool invocation. Stores agent identity, tool, parameters, response, latency, and policy decision.

⚙️ Policy Engine

Declarative YAML rules that evaluate each request in real time. Supports pattern matching, allowlists, risk scoring, and chained logic.

🔔 Alerting

Webhook, Slack, and PagerDuty integrations trigger on block events or custom conditions. SIEM export available on Business+ plans.

Full docs ship with beta access

API reference, policy language spec, SDK guides, and integration docs are all available to beta customers. Join the waitlist to get access.

Join the Waitlist →