Build with ContextGuard

SDKs

@ctg/sdk (TypeScript/JavaScript, npm) and ctg-sdk (Python, PyPI) scan chat messages, protect LLM API calls, and can wrap an OpenAI client so every call is scanned automatically.

npm install @ctg/sdk

import { DlpyClient } from "@ctg/sdk";

const client = new DlpyClient({
  apiKey: "ctg_your_key_here",
  baseUrl: "https://your-ctg-gateway.example.com",
});

const result = await client.scanText("My SSN is 123-45-6789");

REST API

Authenticate with an x-user-id header (your Entra Object ID) against your ContextGuard gateway deployment.

• POST /v1/guardrail — scan a list of chat messages against active DLP policies; returns the (possibly redacted) content, any violations, and whether the request was blocked.
• POST /v1/chat/completions and POST /v1/completions — OpenAI-compatible proxy; requests and responses are scanned before/after forwarding to the model.

Full schema: OpenAPI spec.

MCP

This site runs a read-only MCP server at contextguard.dev/mcp (Streamable HTTP transport, no auth) with tools for looking up the product, pricing, and docs — manifest at /.well-known/mcp.json.

The product itself also runs an MCP Gateway: point an MCP client at /mcp on your ContextGuard gateway deployment to reach every enabled upstream. Tool calls are inspected and governed by the same DLP policies that cover chat traffic.

Webhooks

Policies can be configured to fire outbound alert webhooks when a violation is detected, alongside the dashboard and audit log — configure endpoints from the console.

More

• llms.txt — when to use ContextGuard and where the rest of this page lives, in a format agents can read directly.
• sitemap.xml — every page on this site.