Generating payslips is a task; understanding where labour cost sits and how it moves is the work that gets skipped because assembling the numbers takes an afternoon. Putting the salary-group figures and the cost dashboards behind a conversation means that analysis happens alongside the routine operations rather than instead of them. Bear in mind what the token grants: this reaches real payroll data and can send real payslips, so where you store it matters as much as any other credential.
An MCP server for payroll management on DingTalk's smart compensation platform. It opens the payslip lifecycle and the reporting behind it to an assistant, so HR work that meant logging in, querying and assembling reports by hand can be done by asking — including the analysis that normally comes after the export.
- Salary groups found by name, and one group's detail: headcount on payroll, labour cost, gross and net pay
- Payroll accounting exports — the individual income tax table and the bank table
- Payslip operations: generate payslips from a salary group, upload via Excel, and send the current month's payslips
- Live send-and-confirm status, so you can see which payslips have been received and acknowledged
- Reporting: labour cost statistics, a compensation dashboard and a cost dashboard
- Configuration through `npx`, with the environment and access token supplied in the client entry
You must already be a user of DingTalk's smart compensation product — this reaches your existing account rather than standing alone. The client entry runs `npx` against `@rlj/salary-mcp@latest` with two environment values: the environment selector and the access token that authenticates you; the README's sample token is an example and must be replaced with your own. Configuration examples are given for Cursor and Cline. The npm package name is `@rlj/salary-mcp` (0.0.28), with the `mcp-server-salary-web` binary.
One command plus a key — npx -y @rlj/salary-mcp@latest, then supply credentials