How to Connect HiBob to Cursor via MCP

💡 No install? Use cloud.anythingmcp.com directly. Sign in, click Connectors → HiBob, paste your credentials, mint an MCP API key and you're done. No Docker, no git clone, no local server.

HiBob + Cursor

The HiBob (Bob) HR platform through its public API with a service user: employee search and records, work, employment, lifecycle and salary history, custom tables, company lists, time off with who is out, requests, balances and policies, holidays, saved reports, onboarding, tasks and documents. Master data writes install switched off.

Prerequisites

See the full setup instructions baked into the connector (visible in the in-app store when you select the connector). The required environment variables for this connector are:

HIBOB_SERVICE_USER_ID, HIBOB_SERVICE_USER_TOKEN

Step 1 — Get credentials

Setting up takes about 10 minutes and needs a Bob admin. The connector uses Bob's public REST API (https://api.hibob.com/v1) with a service user.

1. In Bob, as an admin, go to Settings > Integrations > Automation > Service Users, click Manage and create a service user with a name and a display name. Copy its ID and token right away: the token is shown only when it is generated. A lost token can be regenerated, which invalidates the old one.
2. Create a dedicated permission group for the service user and add it there. A new group has no permissions, so grant only what you use:
   • People's data > People's fields: View on the categories you read (at least root, about, employment and work for hibob_search_people), View history on the history tables (work, employment, lifecycle, payroll), Edit where you update.
   • Access data for: Everyone, or a condition. By default only employed people are covered; remove the condition Lifecycle status equals Employed to read leavers with show_inactive.
   • People's data > Time off: view requests and balances, and submit or cancel requests if you use those tools; view private requests for include_private.
   • Features > Reports: View reports according to people's data access rights.
   • People's data > Tasks and flow > People's tasks: View and mark selected people's tasks as complete.
   • People's data > Docs > [folder]: View per document folder, and Upload for the shared folder.
3. Paste the ID into HIBOB_SERVICE_USER_ID and the token into HIBOB_SERVICE_USER_TOKEN. They are sent as HTTP Basic auth.
4. Install the connector and run hibob_list_employee_fields to check the access.

Field ids first: hibob_list_employee_fields lists every built-in and custom field with its id in dot notation (root.displayName, work.department, work.custom.field_123) and the list id of list fields. hibob_list_named_lists and hibob_get_named_list turn list ids into names (departments, sites, job titles, calendars).

What it reads:

• People: hibob_search_people reads many employees in one call (everyone, or filtered by root.id or root.email only; other filters are refused), hibob_get_employee one person by id or work email, hibob_list_profiles a quick directory. For a large company the AI first searches with fields: ["root.id"] and then reads 50 to 200 ids per call.
• History tables for one person (work, employment, lifecycle, salary) and in bulk for many (work, lifecycle, salary).
• Custom tables with their columns and the entries of one employee.
• Time off: who is out in a date range or on one day, requests changed since a time, one request with its days, an employee's balance for a policy type, policy types, reason codes, policies, and company holidays and closures with hibob_search_calendar_events.
• Saved reports, run as JSON rows, CSV or XLSX; large ones are generated in the background with hibob_request_report_download and fetched with hibob_download_report_by_name.
• Onboarding wizards, open tasks across the company, the tasks of one employee, document folders and an employee's documents with download links.

What it writes when installed: hibob_submit_time_off_request (after the AI checks the policy type's unit with hibob_get_policy_type; the request follows the policy's approval flow unless an admin sets skip_manager_approval), hibob_cancel_time_off_request and hibob_complete_task.

Switched off at install: the master data writes hibob_update_employee (fields only, not table rows), hibob_create_work_entry (promotion, transfer, new manager), hibob_create_custom_table_entry, hibob_add_named_list_item and hibob_upload_shared_document (a public or pre-signed URL, at most 5 MB, visible to the employee). A workspace admin turns them on deliberately in the connector.

Data formats: employee ids are long numbers sent as strings (3332883884017713238). Requests use dot notation (work.department); answers use slash keys (/work/department) and may also nest the same value by category. List fields hold item ids; human_readable: APPEND or REPLACE adds or puts the names. Dates are YYYY-MM-DD; hibob_list_time_off_changes takes ISO 8601 date-times. Balances are in the policy type's unit (days or hours), and salaries carry their own currency. After an update Bob may return the old value for up to 20 seconds.

Silent omission: a field without permission or with a wrong id is simply missing from the answer, with no error. When something is missing, check the service user's permission group before concluding the value is empty.

Paging: people search, profiles, the history of one employee, open tasks (at most 5,000) and who's out return everything at once. The bulk history tools and hibob_search_calendar_events page with cursor set to response_metadata.next_cursor; the AI stops when it is null.

Limits: rate limits apply per endpoint (people search allows 50 requests per minute) and answer 429. Who's out and calendar ranges have maximum spans, and time off changes reach back at most 6 months.

Common errors: 401 means a wrong ID or token, 403 a missing permission in the service user's group, 400 an invalid filter, field or date range. Repeated 401 or 403 answers can be blocked for a while by Bob's firewall, so fix the permissions before retrying.

The connector does not run payroll, does not change history table rows through hibob_update_employee, and has no tool to approve or decline pending time off requests.

Step 2 — Install the adapter

curl -fsSL https://raw.githubusercontent.com/HelpCode-ai/anythingmcp/main/docker-compose.quickstart.yml -o docker-compose.yml
printf 'JWT_SECRET=%s\nENCRYPTION_KEY=%s\n' "$(openssl rand -hex 32)" "$(openssl rand -hex 32)" > .env
docker compose up -d

Step 3 — Add the connector in Cursor

Your server URL: in AnythingMCP, open MCP Servers → the server this connector is on, and copy its URL (https://cloud.anythingmcp.com/mcp/…). Use it wherever this guide shows YOUR_SERVER_ID.

Cursor reads MCP servers from ~/.cursor/mcp.json. Add this entry:

{
  "mcpServers": {
    "anythingmcp": {
      "url": "https://cloud.anythingmcp.com/mcp/YOUR_SERVER_ID",
      "headers": { "X-API-Key": "YOUR_MCP_API_KEY" }
    }
  }
}

1. Create an MCP API key in AnythingMCP under MCP Servers → your server → API keys, and paste it in place of YOUR_MCP_API_KEY.
2. Save the file and restart Cursor.
3. Open Cursor → Settings → MCP and check that anythingmcp is listed and connected.
4. Start chatting: the agent can now call the HiBob tools.

Available tools

Example prompts

• "Who is out of the office next week, and with which time off policy?"
• "Show the work history of [email protected]: titles, departments and managers with their effective dates."
• "How many holiday days does employee 3332883884017713238 have left this year, and how many are pending?"
• "Run the saved report "Headcount by department" and summarise it by site."
• "List all open onboarding tasks that are overdue, with owner and the new hire they belong to."

FAQ

Does Cursor support custom MCP servers on the free tier? Yes, MCP is available on Hobby, Pro and Business plans.

Next steps

• Connector landing page
• Connect to Claude
• Connect to ChatGPT
• Connect to Gemini
• Connect to Meta Muse
• Connect to Microsoft Copilot