Skip to main content

Agent Connectors

Connect autonomous AI agents to CVD Portal to answer compliance questions and track deadlines. This guide covers setup for Claude, OpenAI custom GPT Actions, and the OpenAI Responses API.

Prerequisites​

  1. An active CVD Portal account on the Reporting plan or higher.
  2. A Bearer API key generated in your dashboard under Settings and API Keys.
  3. The read scope assigned to the API key.

Claude remote MCP connector​

Claude Desktop and Claude Code can connect directly to the CVD Portal remote Model Context Protocol (MCP) server.

Configuration​

Add CVD Portal to your Claude configuration file.

{
  "mcpServers": {
    "cvd-portal": {
      "url": "https://cvdportal.com/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Supported read queries​

Once connected, you can ask questions directly in Claude.

  • Ask CVD Portal how long we have to report.
  • When is our next CRA Article 14 deadline?
  • What is our next CRA compliance task?

Claude calls get_next_deadline and get_next_task to return accurate regulatory data.

OpenAI custom GPT Action​

You can connect a custom GPT in ChatGPT to CVD Portal using the OpenAPI specification.

Setup instructions​

  1. Open ChatGPT and create or edit a custom GPT.
  2. Under the Configure tab, select Create new action.
  3. In the Schema field, choose Import from URL.
  4. Enter the schema URL https://cvdportal.com/openapi.json.
  5. Under Authentication, select API Key and choose Bearer format.
  6. Paste your CVD Portal API key.
  7. Save your action.

Your custom GPT can now query /compliance/next-deadline and /compliance/next-task.

OpenAI Agents SDK and Responses API​

You can connect the OpenAI Responses API or the Agents SDK to CVD Portal through remote MCP.

Example code​

import { Agent } from "@openai/agents"

const complianceAgent = new Agent({
  name: "CRA Compliance Assistant",
  instructions: "Help the team track CRA deadlines and obligations using CVD Portal.",
  mcpServers: [
    {
      name: "cvd-portal",
      url: "https://cvdportal.com/api/mcp",
      headers: {
        Authorization: `Bearer ${process.env.CVD_PORTAL_API_KEY}`,
      },
    },
  ],
})

The agent uses get_next_deadline and get_next_task as tools during conversation turns.