Set Up the Happeo MCP Server

🎯 Who this article is for: Happeo admins complete the initial setup. An organization admin in your AI assistant registers the connector. All users can then connect their own accounts and start using Happeo from within their AI assistant.

💼 Package requirements: Available to Starter, Growth, and Enterprise. Legacy packages, including recommended and custom, are not entitled to Happeo MCP. Reporting a knowledge gap through your AI assistant additionally requires the Knowledge Engine add-on.

🔒 Permissions: Creating the OAuth client requires Happeo admin access. Registering the connector in Claude, for example, requires Claude organization admin access. Individual users can connect their own accounts once the admin setup is complete.

💻📱Devices: Desktop and mobile, through your AI assistant's own apps.

1. Overview

Happeo MCP is a secure connection between your intranet and your team's AI assistant. Once it is set up, someone can ask their assistant "what is our expense policy?" and get the answer from your company's own documentation.

Model Context Protocol (MCP) is the open standard that makes this possible. It is a common way for AI assistants to connect to the systems a company already uses, and no single AI company owns it, which is why the same connection can reach more than one assistant over time. Happeo MCP is available today with Claude. ChatGPT and Gemini are coming next.

Happeo MCP inherits each user's existing Happeo permissions in full. Whatever someone can read or write in Happeo, they can read or write through their AI assistant. Whatever they cannot access in Happeo, they cannot access through the assistant either.

2. Use cases

• Ask a question and get an answer from your own documentation: Someone working in their AI assistant asks what the parental leave policy is, who leads a team, or how to request new software. The assistant answers from the Happeo content that person is allowed to see and lists its sources, and follow-up questions continue against the same context, so a general policy question can become a specific edge case without starting over.
• Catch up on what changed, personalized to each person: Someone asks their assistant to catch them up on the updates and conversations relevant to them this week. It synthesizes across the Pages, channel activity, and notifications that person can access, and returns what changed, what matters to them, and what needs their attention.
• Turn company context into finished work: Ahead of a customer call or an all-hands, someone asks their assistant to prepare a briefing using what the company already knows. It pulls the relevant Pages, channel activity, and People Directory entries, and returns the finished piece with its sources listed, instead of the person assembling that context by hand.
• Publish useful knowledge back into Happeo: Work does not have to stop at a draft. Someone can ask their assistant to publish a finished update as a post in a Happeo channel, or to publish a comment on a post they have just read, so the answer lands where the rest of the team already looks for it.

3. Before you begin

Before starting, confirm the following:

• A Happeo admin is available to create the connection in Happeo.
• Someone with admin access to your organization's AI assistant account is available to register Happeo in that tool.
• Anyone who wants to connect will need their own account with the AI assistant you're using (for example, a Claude account).

4. How to set up the Happeo MCP server

5. What Happeo MCP can do

Everything below happens by asking in plain language, and everything is scoped to what that person can already see in Happeo.

• Answer a question from your intranet: Someone asks a question the way they would say it out loud, for example "what is our parental leave policy?". The assistant searches the Pages, Channels, and People Directory that person can access, answers from what it finds, and lists what it used under "Sources:", with each title hyperlinked so anyone can check the original in Happeo.
• Keep going with follow-up questions: The conversation carries on against the same Happeo context. After a policy answer, someone can ask "what if the employee is based outside Finland?" and get a further answer without starting a new search.
• Look up a colleague: Someone asks who leads a team or owns a topic, and gets the answer from your People Directory.
• Catch someone up on what changed, personalized to them: Someone asks something like "catch me up on the updates and conversations relevant to me this week", and gets one view of the important updates, the changes worth knowing about, and anything waiting on them. It is a synthesis across everything that person can see rather than a summary of one channel, which is why two people asking the same question get different answers.
• Summarize one channel: Someone names a channel and gets a summary of its latest posts and comments.
• Surface the notifications that matter: Someone asks what has changed for them, and the relevant notifications come into the conversation they are already having.
• Produce finished work from company context: Someone asks for a customer briefing, an all-hands update, or a first draft. The assistant pulls the relevant Pages, channel activity, People Directory entries, and policies, and returns the finished piece with its sources listed.
• Publish a post to a channel: Someone can ask the assistant to publish something it has drafted as a post in a named Happeo channel. It appears exactly as it would if the person had written it in Happeo directly.
• Publish a comment: Someone can ask the assistant to publish their reply as a comment on a post, so a response written in the assistant lands where the team sees it.
• Search your connected Enterprise Search connectors: If your organization has Enterprise Search connectors configured, for example Google Drive, Jira, or HubSpot, the assistant returns results from those tools alongside Happeo content, still limited to what that person can see. Search and retrieval only: it cannot take an action in those tools, such as closing a Jira ticket.
• Report a knowledge gap: Someone who already knows something is missing or out of date can ask the assistant to log it as a content gap. It lands on the same Gap Dashboard your admins already use. The assistant never logs a gap on its own. Requires the Knowledge Engine add-on.

6. What Happeo MCP cannot do today

These are boundaries rather than settings, so they are worth knowing before you roll Happeo MCP out.

• Publishing covers posts and comments, and nothing else: Articles, announcements, and polls are not supported content types for this connection. Those are still created in Happeo directly.
• One channel per publish: Each request targets a single channel. The same update in three channels is three requests, or one post written in Happeo.
• No posting on behalf of someone else: Every action runs with the permissions of the person who connected their own account. There is no delegated or shared-account posting, so nobody can publish as a colleague or on behalf of a team.
• No creating or editing pages: Creating a page, and publishing an edit to an existing page, are not available through Happeo MCP today. Both are on the roadmap, with no committed date. Pages are still created and published in Happeo directly.
• No admin control to switch Happeo MCP on or off: It comes with your package, and there is no toggle in Admin Settings to enable or disable it. To turn it off for your whole organization, contact Happeo Support.

7. Best practices

A few habits keep the connection secure and useful over time.

• Set an expiration date on your connection password: Rather than selecting Never expires, giving your connection password an end date limits the risk if it is ever lost or exposed. When it is due to expire, generate a new one in Happeo, update the details in your AI assistant, and delete the old one immediately.
• Ask in plain language: The connection is built to respond to natural questions. Encourage your team to ask the AI assistant the same things they would type into Happeo Search: "What is our remote work policy?", "Who works in the Helsinki office?", or "Summarize the latest posts in the kudos channel."
• Test before rolling out: After setup, try a few questions to make sure the AI assistant is returning the right results and respecting access levels before sharing the connection with the wider team.

8. Letting your team know

Once the connection is live and tested, share it with your team. A few things worth covering in your announcement:

• What it does in plain terms: Most people are not familiar with MCP. Skip the technical explanation and describe the outcome: they can now ask their AI assistant questions about the company and get answers from Happeo directly.
• How to get connected: Users need to link their own accounts before they can use it. Tell them where to go (Connectors settings in their AI assistant) and what to expect (a quick sign-in flow).
• What to try first: Give one or two example questions that match something your team actually looks up often. A concrete example is faster than a general explanation.
• What it does not do: Worth setting expectations — the AI assistant only acts when asked, and it can only access what each person already has permission to see in Happeo.

9. Frequently asked questions

10. Troubleshooting