MCP Protocol: Integrating Compliance Into Your AI Workflows

Introduction

Compliance checks shouldn't require a human to click buttons in a web UI. If you're building a product development pipeline, an ERP integration, or an AI assistant for your formulation team, you need compliance as a function call — not a manual step.

Inlet's MCP server exposes the full compliance engine as a set of tools that any MCP-compatible AI agent can invoke. Run a compliance check, search ingredients, query regulations, generate reports — all programmatically, with the same tenant isolation and authentication as the REST API.

What is MCP?

The Model Context Protocol (MCP) is an open standard for connecting AI models to external tools and data sources. Instead of building custom integrations for each AI platform, you implement the MCP server once and any compatible agent can use it.

In practice, MCP lets an AI agent discover what tools are available (via GET /mcp/tools), invoke a tool with structured parameters (via POST /mcp/execute), and receive structured results back. The agent doesn't need to know how Inlet works internally — it just calls the tool and gets a compliance determination.

Available Tools

Inlet's MCP server exposes tools across the full compliance lifecycle:

Execution Modes

Single execution

Call one tool at a time. Ideal for interactive AI assistants that make one request per user turn.

{
  "tool": "compliance.check",
  "parameters": {
    "sku_id": "76349bc1-ef2a-4ac2-94ef-11a4e8d66942",
    "markets": ["US", "EU"]
  }
}

Batch execution

Run multiple tools in a single request. Useful for pipeline automation — check 50 SKUs against 3 markets in one call without 150 round-trips.

{
  "calls": [
    { "tool": "compliance.check", "parameters": { "sku_id": "sku-1", "markets": ["US"] } },
    { "tool": "compliance.check", "parameters": { "sku_id": "sku-2", "markets": ["EU"] } },
    { "tool": "ingredients.search", "parameters": { "query": "Retinol" } }
  ]
}

Streaming

Stream results for long-running operations. The server sends results as they complete rather than waiting for the entire batch to finish.

{
  "tool": "compliance.check",
  "parameters": { "sku_id": "sku-1", "markets": ["US", "EU", "UK", "CA", "ASEAN"] },
  "stream": true
}
// Server-sent events:
// data: {"market": "US", "status": "compliant", "confidence": 0.98}
// data: {"market": "EU", "status": "needs_review", "confidence": 0.82}
// ...

Example Workflows

Product development assistant

Build a chatbot for your formulation team that answers questions like "can I use Retinol at 1% in an EU face cream?" The assistant calls ingredients.search to find Retinol, then regulations.query to check EU restrictions, and returns a grounded answer with the specific regulation citation.

ERP integration

When a new product formulation is saved in your ERP system, trigger a batch compliance check across all target markets. If any market returns "non_compliant", block the formulation and notify the regulatory affairs team with specific violation details.

CI/CD for formulations

Treat compliance like a test suite. Every time an ingredient changes in a product formulation, run the full compliance check battery. If a check fails, the "build" fails — preventing non-compliant formulations from reaching production.

Authentication

MCP requests authenticate via API keys with the same tenant isolation as the REST API. Create an API key in Settings, and include it in the Authorization header. Keys use ink_live_ prefix and are SHA-256 hashed at rest.

Authorization: Bearer ink_live_your_api_key_here

Getting Started

The MCP server is available at https://inlet.run/api/v1/mcp. Start by listing available tools, then try a single compliance check. The full API documentation is at /docs/api in the app.