What is MCP (Model Context Protocol)? It is an open standard that lets an AI app, such as Claude or ChatGPT, connect to other software and use its tools: read data, create records, take actions. Instead of copying and pasting between your AI chat and your project tool, the AI works inside the tool directly. AlianHub ships an MCP server, so the assistant you already use can run real project work for you.
Key takeaways
- MCP is a common plug that connects AI apps to the software where your work lives.
- Without MCP, your AI can only talk about your work. With it, the AI can do the work.
- The software decides what tools it offers. The AI can only use what is on the list.
- In AlianHub, a connected AI acts as you, with your role and permissions, and nothing more.
- Wider changes wait for approval, and every change is logged under its name and yours.
MCP explained simply
Think of MCP like a USB port for AI. Before USB, every device needed its own cable. Before MCP, every AI integration was custom-built.
MCP defines two sides:
- The client is the AI app, for example Claude, ChatGPT or Claude Code.
- The server is the software you want the AI to use. It publishes a list of tools, such as "find tasks" or "add a comment".
When you ask the AI something, it looks at the tools available, picks the right one, and calls it. The server runs the action and returns the result. You see the answer in your chat.
Why it matters for businesses
Most of the value in AI comes from context. A model that cannot see your projects gives generic advice. A model connected through MCP can answer "what is overdue on the launch?" from your real data, and then add the follow-up task.
What a well-designed MCP server should do
Connecting AI to business data raises fair questions about safety. A good server answers them with design, not promises.
| Concern | What good design looks like | How AlianHub does it |
|---|---|---|
| Too much access | AI sees only what the user can see | It acts as you, with your role and permissions |
| Risky actions | Dangerous tools are not offered at all | Deleting tasks, docs, comments or projects is not available |
| Silent changes | Wider changes need a person | Batches, new projects and rules wait in the AI Inbox |
| No trail | Every action is recorded | Logged under the AI's name and yours |
| Hard to switch off | Revocation is instant | Revoke under Accounts, Connected apps |
The safest tool is the one that is not on the list. In AlianHub, the risky actions are absent from the tool list, not merely disabled.
How AlianHub uses MCP
AlianHub's MCP server works with Claude, ChatGPT and Claude Code. Once connected, you can ask in everyday words to find, add and change work.
What it can do out of the box
- Find and read your tasks, projects, lists, docs, comments and time entries.
- Add tasks and subtasks, comment, attach links and log time.
- Change a task's status to In progress or In review.
- Show you where something is, with a link.
With the extra permissions on the consent screen, and unless your admin has switched those tools off, it can also change, assign, move, archive and close tasks, write docs, and read chat in channels you are in. Reading chat is its own permission and never includes direct messages.
What it can never do
It cannot delete a task, a doc, a comment or a project. It cannot remove people, change who may do what, or touch billing. It cannot do more than you can. And it does nothing while its app is closed, unless you set up a schedule in that app.
Tip: open the screen you mean in AlianHub before asking. "Here" and "this list" refer to the screen you last opened.
How to do this in AlianHub
- Sign in to AlianHub. Your account comes from an admin invitation, or from the install wizard if you set up the server yourself.
- Open AI, then Connect your AI, and copy the address shown.
- In Claude, add it as a custom connector and paste the address. In ChatGPT, open Settings, then Apps and Connectors, turn on Developer mode under Advanced settings and create a connector (ChatGPT offers this on some plans only).
- On the consent screen, choose what it may do. Tick the extra permissions only if you need them.
- If your workspace has not approved the app yet, an owner or admin approves it under Settings, Agent clients.
- Back on the Connect your AI page, the status line changes to "Connected". Try the first sentence it suggests: "Set up my project". It asks a few questions and shows a plan; nothing is made until you say yes.
- Review anything that waits in the AI Inbox, and see every change in the Audit log under the "Outside agents" tab.
For Claude Code and other CLI agents, create a personal token under AI, Accounts, My account, scoped to chosen projects. The token is shown once. Your server address must be reachable from outside your network. The full walkthrough is on the Connect your AI page.
A business scenario: an operations manager
An operations manager at a growing services company lives in ChatGPT for writing and planning. Before connecting, she pasted task lists into chat and copied answers back by hand.
After connecting AlianHub over MCP, she asks "show me all my tasks across every project" and gets a link to her Everything screen. She asks it to add "Call supplier" to the Launch list, and it is applied at once with Undo. When she asks it to set twenty tasks to high priority and assign them, the batch waits in the AI Inbox as one request.
She keeps working in the AI app she knows, and the project tool stays the single source of truth. No more copy-paste drift between chat and the plan.
MCP vs built-in AI
AlianHub also has built-in AI: the Ask box, agents with autonomy levels L0 to L3, and skills like digests and brief breakdowns. MCP is for people who want their own assistant to work for them. Agents are for work the team hands off. Many teams use both. See AI agents for project management and the AI page.
FAQ
Is MCP only for Claude?
No. MCP is an open protocol supported by several AI apps. AlianHub's guide covers Claude, ChatGPT and Claude Code.
Can a connected AI see projects I cannot?
No. It sees only what you can open and changes only what your role allows. If you cannot open a project, neither can it.
How do I disconnect an AI app?
Revoke it under AI, Accounts, My account, or under Accounts, Connected apps. It stops on its next call.
Does my data leave my server?
Your AI app reaches AlianHub over the internet and reads what it needs to answer. The data itself stays in your own database on your server.
Ready to try it? Explore the live demo, self-host AlianHub free, and share this explainer with anyone on your team still asking what MCP is.