Dutify 教學：讓 AI 智能體以單一 MCP 連接器接入四項工作區產品

Most AI agents start their first day the way a new hire does: one login for the task tracker, another for the docs, another for call recordings, another for customer feedback. Every extra tool is another token to manage, another set of permissions to reason about, and another place where the agent's picture of the work goes stale.

This tutorial connects an agent to all four through one hosted MCP server and one API key, using Dutify. It takes about five minutes.

What you get

One remote MCP server at https://mcp.dutify.ai (Streamable HTTP), covering four products:

Product	Tools	What they cover
Dutify Cloud (projects)	`pm_*`	Tasks, subtasks, comments, sprints, views, custom fields, time tracking
Dutify Hub (calls)	`hub_*`	Call recordings, transcripts, summaries, search across calls
Codexum (wiki)	`wiki_*`	Spaces, pages, search, attachments
Roadmarq (feedback)	`fr_*`	Feature requests, bugs, votes, statuses

Plus a handful of discovery tools (whoami, list_workspaces, find_my_workspace) so the agent can orient itself. The tool list is public; every data call needs a key, and the backend enforces that key's scopes and your workspace permissions on each call.

1. Get a key

Sign up at dutify.cloud (free), then create a personal key in account settings. It starts with du_live_ and works across every product your account can access.

If you'd rather keep an agent inside one product, workspace keys exist too: dk_live_… for Cloud, Codexum and Roadmarq, dh_live_… for Hub.

2. Connect your client

Claude Code:

claude mcp add --transport http dutify https://mcp.dutify.ai --header "X-API-Key: du_live_YOUR_KEY"

Cursor, Claude Desktop and other JSON-config clients:

{
  "mcpServers": {
    "dutify": {
      "url": "https://mcp.dutify.ai",
      "headers": { "X-API-Key": "du_live_YOUR_KEY" }
    }
  }
}

VS Code (.vscode/mcp.json):

{
  "servers": {
    "dutify": {
      "type": "http",
      "url": "https://mcp.dutify.ai",
      "headers": { "X-API-Key": "du_live_YOUR_KEY" }
    }
  }
}

3. First calls

Ask the agent to run whoami and then list_workspaces. With a personal key, the agent passes the workspace on each data call; you can also pin one with an X-Dutify-Workspace header.

From there the useful prompts are the ones that cross tools, because that's the part a single-product integration can't do:

"What did we decide about pricing on last week's calls, and is there a task for it?"
"Summarize the open feature requests tagged Jira and link any related wiki pages."
"Create a task for each action item from today's call and assign it to the person who took it."

A note on scope

One connector does not mean one giant permission. A personal key can only reach what your account can reach, and a workspace key is bound to one workspace and product family. For an agent that only needs to read the roadmap, give it a narrow key, not your personal one.