MCP Server — Docs

Premium Ultimate

Unicis Platform ships an open-source MCP (Model Context Protocol) server that connects AI assistants such as Claude, Cursor, and VS Code Copilot directly to your Unicis data. Manage tasks, privacy records, cybersecurity controls, risk registers, and file attachments through natural language.

The server lives at src/mcp-server inside the unicis-platform monorepo. It’s a standalone Node.js package with its own package.json/tsconfig.json — it talks to the platform over its public REST API and does not share a build with the Next.js app.

Prerequisites

• Node.js 18+
• A Unicis Platform API key — generate one at Team Settings → API Keys (see Settings — API Keys)

Install and build (local)

Clone the platform monorepo and build the MCP server package:

git clone https://github.com/UnicisTech/unicis-platform-ce.git
cd unicis-platform-ce/src/mcp-server
npm install
npm run build

This produces dist/index.js, the server entry point.

Environment variables

Mode 1 — Claude Desktop (local, stdio)

This is the simplest way to run the server locally on your own machine, with Claude Desktop launching it as a subprocess.

Edit your Claude Desktop config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "unicis": {
      "command": "node",
      "args": ["/absolute/path/to/unicis-platform-ce/src/mcp-server/dist/index.js"],
      "env": {
        "API_BEARER_TOKEN": "your-api-key",
        "API_BASE_URL": "https://platform.unicis.tech"
      }
    }
  }
}

Use the absolute path to dist/index.js from the build step above. Restart Claude Desktop — all unicis_* tools appear automatically, no further configuration needed.

Mode 2 — Remote HTTP server (claude.ai web, Cursor, VS Code)

To make the server reachable by clients that can’t spawn a local subprocess (the claude.ai web app, Cursor, VS Code Copilot, or any teammate’s machine), run it in HTTP mode on a host you control — a VPS, container, or internal server:

TRANSPORT=http \
API_BEARER_TOKEN=your-api-key \
API_BASE_URL=https://platform.unicis.tech \
npm start

Then point your MCP client at the running endpoint. For Claude Desktop connecting to a remote endpoint via mcp-remote:

{
  "mcpServers": {
    "unicis": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.unicis.tech/mcp",
        "--header",
        "Authorization: Bearer your-api-key"
      ]
    }
  }
}

Verifying your setup with the MCP Inspector

Before wiring the server into a client, you can inspect and call its tools directly:

API_BEARER_TOKEN=your-key API_BASE_URL=https://platform.unicis.tech \
  npx @modelcontextprotocol/inspector node dist/index.js

This opens an interactive UI where you can list the registered unicis_* tools and test calls against your team’s data before connecting a real AI client.

Available tools

Tasks

Status values: todo · inprogress · done. Priority values: low · medium · high.

Privacy — RoPA, TIA, PIA

Cybersecurity Controls (CSC)

Supported frameworks: ISO/IEC 27001 (2022 & 2013), MVSP, NIST CSF v2, EU NIS2, GDPR, CIS CSC v8.1, SOC 2 v2, C5 2020, OWASP ASVS v5, PCI DSS v4.0.1, ISO/IEC 42001.

Risk Management

API Keys

Not yet supported

The IAP (Interactive Awareness Training) module — course assignments, completion tracking, quiz results, learner reports — does not currently have MCP tool coverage. Planned for a future release.

Development

npm install
npm run build

Project structure:

src/
├── index.ts            # Server entry point — registers all tools, sets up transports
├── services/
│   └── api.ts          # Typed API client (apiGet / apiPost / apiPut / apiDelete)
└── tools/
    ├── tasks.ts         # Task CRUD, comments, attachments, controls
    ├── compliance.ts    # CSC statuses, API key management
    ├── privacy.ts       # RoPA, TIA, PIA tools
    └── risk.ts          # Risk management tools

Contributions are welcome — see the platform’s CONTRIBUTING guide. New tools must include typed Zod input schemas and match the Unicis Platform REST API exactly; open a GitHub issue first for significant new features.

Multi-team / MSSP setup

API keys — and therefore MCP server connections — are scoped to one team. If you’re an MSSP or consultancy managing several client tenants, register one mcpServers entry per team, each with its own API key:

{
  "mcpServers": {
    "unicis-client-a": {
      "command": "node",
      "args": ["/absolute/path/to/unicis-platform-ce/src/mcp-server/dist/index.js"],
      "env": {
        "API_BEARER_TOKEN": "client-a-api-key",
        "API_BASE_URL": "https://platform.unicis.tech"
      }
    },
    "unicis-client-b": {
      "command": "node",
      "args": ["/absolute/path/to/unicis-platform-ce/src/mcp-server/dist/index.js"],
      "env": {
        "API_BEARER_TOKEN": "client-b-api-key",
        "API_BASE_URL": "https://platform.unicis.tech"
      }
    }
  }
}

Each entry runs as its own server process but shares the same dist/index.js build, so there’s nothing extra to install per client — only a new API key and a new named entry. Your AI assistant then has all configured teams available side by side in the same conversation, so you can manage tasks, controls, and risk records across every client tenant from one place instead of switching context per team.

Why it matters

Compliance work is full of repetitive, multi-step tasks: pulling evidence, updating statuses, drafting documentation, following up on overdue items. The MCP server lets you hand these off to an AI assistant that can act directly on your platform data, turning a chat into a working session instead of a research exercise.

Availability

See Pricing for full plan details.

Additional resources

• GitHub repository (src/mcp-server)
• Unicis Platform REST API (Swagger UI)
• MCP specification
• Model Context Protocol SDK (TypeScript)