HR MCP server
Use your Skailer data in Claude, ChatGPT, Cursor and VS Code, with your own permissions
01Signs in as you
Each person connects their own account. The client sees what their role opens, not an admin's view.02Read-only if you choose
Clear "Create and update data" on the access screen, and the client only reads.03Checks before it creates
A time off request is checked for balance and approvers before it's created.04Nothing gets deleted
The connector can create and update a short list of things. It has no delete actions.The problem
HR data shouldn't reach your AI chat as a pasted spreadsheet
Your team already drafts, plans and codes in Claude, ChatGPT, Cursor or VS Code. When a question needs HR data (who is off during the release, which goals slipped, what is waiting for approval), someone exports a table and pastes it in. The export is out of date by the next day, and it shows the whole table rather than what that person may see. The alternative is a connector on an admin API key, which gives the AI client everyone's data. Requesting leave or approving a request still means switching to the HR system. Skailer changes that by letting each person connect their own account, so the AI client works with live data under that person's permissions.
Connect your client
Connect Claude, ChatGPT, Cursor or VS Code in a few minutes
Add the Skailer server address in your client, sign in to Skailer in the browser, choose what the client may do and click Allow. Claude Team and Enterprise owners add the connector once for the whole team, and each member connects their own account. Developers add Skailer to Claude Code with one command or to Cursor and VS Code through mcp.json.
01Claude
Free, Pro and Max as a custom connector; Team and Enterprise through Organization settings; Claude Code with one command.02ChatGPT
Through developer mode. Business, Enterprise and Edu read and change data; Pro reads only.03Cursor
In mcp.json, for all projects or one.04VS Code
In GitHub Copilot chat through MCP: Add Server or mcp.json.05Other clients
Any client that adds a remote MCP server by address and signs in through the browser.Ask about your company
Ask about people, leave, goals and reviews without leaving your chat
Once connected, the client reads the Skailer data you're allowed to see and answers with it. Ask who is off during a release week, which goals in your team are at risk, how a review cycle is going or what is waiting for you. The answer comes from live data, so nobody exports a file first.
01People
Search employees, open a profile, find someone by email.02Time off
Balances, categories, team and company leave calendars.03Goals
Search goals and open one with its key results.04Surveys and reviews
Results, answers, participants and review stats.05Things to do
Reviews, surveys, tasks, documents to sign, candidate feedback and approvals, overdue first.Act from the chat
Request leave, approve requests and update goals from the chat
If you allowed "Create and update data available to you", the client can also make changes, the same short list SK.AI can make in Skailer. A time off request is checked first: your balance on the first day off, the balance left and who approves. The client creates the request after you agree, and it goes to approvers as usual. If it breaks a policy rule, nothing is created and the client says why.
01Goals
Create and update goals and key results, add check-ins.02Your time off
Request and cancel your own requests.03Approvals
Approve or decline time off waiting for you, with a comment.04Recognitions
Give a colleague a recognition you're allowed to give.05Stays in Skailer
Offers, position requests and other approvals are listed, and you decide on them in Skailer.Your permissions, not an admin key
Give every connection one person's access so nothing is over-shared
The connector has no organization-wide key. Each person signs in to Skailer from their client, and the client gets exactly what that person's role opens. Reading is always included; changing data is a separate permission the person can leave unchecked. A permission added later needs a reconnect, and removing the connector in the client ends access. A new hire whose portal is still limited to onboarding can't connect until the portal opens fully.
01Sign-in per person
OAuth in the browser, no shared token.02Two permissions
View data available to you (always), create and update data available to you (optional).03Role-based scope
The same field and data permissions as in Skailer.04Safe return address
After Allow, Skailer returns only to an https address or localhost.05Disconnect any time
Remove the connector, and the client stops working with Skailer.What the connector can doThe connector exposes 33 tools across seven areas. Read tools work with the default permission. Write tools need "Create and update data available to you". Not in the connector: recruiting and candidates, compensation, documents, settings, delete actions.
See every tool and whether it reads or writes
| Area | Read | Write (needs permission) |
|---|---|---|
| People | Search employees, get an employee, find by email, get me | — |
| Things to do | Get my inbox | — |
| Goals | Search goals, get a goal | Create and update goals, create and update key results, check in |
| Time off | Categories, balance, employee calendar, search time off, team calendar, check a request | Create, cancel, approve and reject requests |
| Surveys | Search surveys, results, answers, participants | — |
| Performance reviews | Search reviews, results, stats, participants, participant report | — |
| Recognition | Search recognitions | Give a recognition |
Try asking
01"Who on my team is off next week?"
02"What do I have to do this week? Overdue first."
03"Which goals in the Sales department are at risk?"
04"Request time off November 10–14."
05"Approve Anna's time off request and add 'Have a good break'."
06"Summarize the open answers in the Q3 pulse survey."
07"Give Alex a recognition for the release."
One HR platform
Pick the way in that fits the job
Inside Skailer, SK.AI answers and acts from the side panel. In Claude, ChatGPT, Cursor or VS Code, the MCP connector does the same under the same permissions. For systems rather than people, the public API and webhooks connect through service users with their own roles.
What is the Skailer MCP connector?
Security
Know what the client can see and where the data goes
The client works only with the data the signed-in person may see in Skailer, and it changes data only if that person allowed it. Choose the client and plan your company already approves for work data.
Skailer vs. the alternativesExports pasted into a chat are the most common route and need no setup, but they show everything in the file and go stale. Connectors built on one integration key give live data, and the key often sees more than any single user should. The Skailer connector signs in as each person.
Three ways HR data gets into an AI chat
| Capability | Exports pasted into a chat | Connector on an integration key | Skailer MCP |
|---|---|---|---|
| Data freshness | As of the export | Live | Live |
| Who the client acts as | Whoever has the file | The key's account, often admin-level | The person who connected |
| Permissions | None after export | The key's scope | The person's role in Skailer |
| Changing data | Not possible | Whatever the key allows | Goals, own time off, approvals, recognitions, only if granted |
| Setup | None | Admin creates and stores a key | Each person signs in once |
| Ending access | Delete every copy | Revoke the key for everyone | Remove the connector in the client |
How it works
How to connect Skailer to your AI client
01Add the server
Paste the Skailer server address into Claude, ChatGPT, Cursor or VS Code. In Claude Team and Enterprise, the owner adds it once for everyone.02Sign in to Skailer
The client opens Skailer in your browser. Sign in as yourself.03Choose permissions
Reading is always on. Keep or clear "Create and update data available to you", then click Allow.04Ask
Turn Skailer on in the chat and ask about your people, leave, goals, reviews and surveys.Skailer MCP: common questions
What is MCP?
MCP (Model Context Protocol) is an open standard that lets AI clients such as Claude and ChatGPT connect to other systems. The Skailer MCP connector is the Skailer side of that connection.