Everjust Payroll Hr

EVERJUST Payroll & HR — Agent Skill

Operate the Payroll & HR app of a live everjust.app tenant as an agent: read employees and their contract "versions", pull real attendance (clock in/out), generate and compute payslips, move a payslip through its state machine, and reconcile the payslip's Worked Days against actual hours. Everything runs through the platform's Odoo MCP (search / get / find / count / create / update / call / describe_model) — see [[everjust-agent-mcp]] for opening the session against the right tenant DB, and [[everjust-platform]] for the platform invariants (one-DB-per-tenant, sudo/ACL boundaries, Odoo-19 API surface) you must not break.

This is mostly stock/OCA: core hr + hr_attendance + hr_holidays, and the OCA payroll module (hr.payslip / hr.salary.rule / hr.payslip.run — NOT the Odoo Enterprise hr_payroll, which this fork does not have). The everjust twist is a single small bridge, everjust_payroll_attendance.

Install state — CHECK THIS FIRST (biggest gotcha)

The prompt-era assumption "not installed on either tenant" is stale. Reality (verified live, 2026-07): install state is uneven across tenants, so before any recipe here, confirm the module is actually installed on the DB you're on:

# via MCP: find on ir.module.module, or just describe_model('hr.payslip')
find("ir.module.module",
     [["name","in",["payroll","hr_attendance","everjust_payroll_attendance"]]],
     ["name","state"])

Observed:

Consequences you must respect:

• everjust_payroll_attendance is auto_install=True and its depends include hr_attendance (plus payroll, hr_work_entry). It therefore installs itself ONLY once hr_attendance is also present. On the four "payroll-but-no-attendance" tenants the bridge is dormant — payslips there use the stock calendar-based _compute_worked_days, not attendance.
• On headsup the bridge IS in the live hr.payslip MRO (verified — the override's source is the one that runs). But headsup currently has 0 salary structures, 0 salary rules, 0 attendances, 0 payslips: the app is installed but unconfigured and un-operated. Computing a payslip with no structure yields an empty payslip (see Pitfalls). Treat "installed" ≠ "ready to pay".
• If the task targets a tenant where payroll isn't installed, stop — installing a module is platform-ops (add to the deploy manifest + -u), not an ORM call you make from here. See [[everjust-platform]].

Key models (real _names and the fields that matter)

Odoo-19 renamed the contract model: a "contract" is an hr.version, not the old hr.contract. Every payroll contract_id (on payslip, worked-days line, and elsewhere) points to hr.version. This is the #1 field-name trap.

hr.attendance and account.analytic.line are independent here. This tenant has no hr_attendance_timesheet-style bridge, so clocking in does not create an analytic line, and the payroll bridge reads hr.attendance, not timesheets. Analytic lines are the Projects/timesheet world — see [[everjust-projects]]. Don't reconcile payroll worked days against account.analytic.line; reconcile against hr.attendance.worked_hours.

How the everjust bridge changes payroll (the one thing that's non-stock)

everjust_payroll_attendance/models/hr_payslip.py inherits hr.payslip and overrides _compute_worked_days(contract, day_from, day_to). During compute_sheet the OCA flow calls get_worked_day_lines → _compute_worked_days; the override:

1. Searches hr.attendance for the employee where check_in is within [day_from, day_to] and check_out != False (open/un-clocked-out spans are ignored).
2. If no such attendances → super() → stock calendar-based worked days (so salaried employees without a clock are unaffected). This fall-through is why the four attendance-less tenants behave stock even though they run payroll.
3. If attendances exist → returns ONE worked-days dict: name="Attendance", code="ATTN", sequence=1, number_of_days = count of DISTINCT check-in dates, number_of_hours = sum(worked_hours), contract_id = contract.id.

So on headsup a computed payslip's worked_days_line_ids will contain an ATTN line whose hours equal real clocked hours and whose days equal distinct clock-in days — that is the everjust behavior to expect and to explain. The money lines (line_ids) still come from the salary structure's rules multiplying against these worked days.

Recipes

Route each through the Odoo MCP against the tenant DB (see [[everjust-agent-mcp]]). create/update/call are shown as MCP verbs; call invokes a model method.

1. Look up an employee, their contract, and clocked attendance for a period

# Find the employee
emp = find("hr.employee", [["name","ilike","Jordan"]],
           ["name","work_email","department_id","job_id","hourly_cost",
            "version_id","resource_calendar_id"], limit=5)
# The contract is the current hr.version (NOT hr.contract):
get("hr.version", emp[0]["version_id"][0],
    ["struct_id","schedule_pay","wage","resource_calendar_id","date_start"])
# Real clocked spans in July 2026 (closed spans only — the bridge ignores open ones):
find("hr.attendance",
     [["employee_id","=",emp[0]["id"]],
      ["check_in",">=","2026-07-01 00:00:00"],
      ["check_in","<=","2026-07-31 23:59:59"],
      ["check_out","!=",false]],
     ["check_in","check_out","worked_hours","in_mode"], order="check_in")
# What the payslip WILL show as worked days (mirror the bridge yourself):
#   number_of_hours = sum(worked_hours); number_of_days = distinct check_in dates.

2. Generate a payroll batch and compute everyone's slip (the correct path)

Do NOT hand-build payslips row by row. Create a run, then call the OCA by-employees wizard, which builds each slip via get_payslip_vals (this is where the attendance bridge fires) and computes them.

# Preconditions: a salary structure with rules must exist, else slips compute empty.
struct = find("hr.payroll.structure", [], ["name"], limit=1)     # must be non-empty!

run = create("hr.payslip.run", {
    "name": "July 2026 Payroll",
    "date_start": "2026-07-01",
    "date_end": "2026-07-31",
    "struct_id": struct[0]["id"],
})

# Build + compute slips for chosen employees, in the run's context (active_id = run):
emp_ids = [e["id"] for e in find("hr.employee", [["active","=",true]], ["id"])]
wiz = create("hr.payslip.employees", {"employee_ids": [[6,0,emp_ids]]})
call("hr.payslip.employees", "compute_sheet", [[wiz]], {"context": {"active_id": run["id"]}})

# Inspect results — each slip now has worked-days lines (ATTN if attendance existed):
slips = find("hr.payslip", [["payslip_run_id","=",run["id"]]],
             ["employee_id","state","number","date_from","date_to"])
get("hr.payslip", slips[0]["id"], ["worked_days_line_ids","line_ids"])
find("hr.payslip.worked_days", [["payslip_id","=",slips[0]["id"]]],
     ["code","name","number_of_days","number_of_hours"])   # look for code == "ATTN"

3. Compute / confirm / cancel a single payslip (its state machine)

hr.payslip.state: draft → verify → done, or → cancel. compute_sheet is what moves draft → verify AND (re)builds all lines; action_payslip_done re-computes (unless prevent_compute_on_confirm) then locks to done.

# From a draft slip: compute it (draft -> verify, fills worked_days + money lines)
call("hr.payslip", "compute_sheet", [[slip_id]])
get("hr.payslip", slip_id, ["state","number","line_ids"])       # state == "verify"

# Confirm (verify/draft -> done). Recomputes first unless prevent_compute_on_confirm.
call("hr.payslip", "action_payslip_done", [[slip_id]])          # state == "done"

# Reset a rejected slip back to draft (ONLY works from state == "cancel"):
call("hr.payslip", "action_payslip_draft", [[slip_id]])

# Cancel. A "done" slip can only be cancelled if the payroll.allow_cancel_payslips
# ir.config_parameter is set; otherwise this raises UserError.
call("hr.payslip", "action_payslip_cancel", [[slip_id]])

For bulk state changes across many slips, the OCA wizard hr.payslip.change.state (change_state_confirm, state ∈ draft/verify/done/cancel, with active_ids in context) applies the same guarded transitions to a selection.

4. Explain a payslip's Worked Days (attendance vs. calendar reconciliation)

# Pull the worked-days lines and compare to raw attendance for the same window:
wd = find("hr.payslip.worked_days", [["payslip_id","=",slip_id]],
          ["code","name","number_of_days","number_of_hours"])
slip = get("hr.payslip", slip_id, ["employee_id","date_from","date_to"])
att = find("hr.attendance",
     [["employee_id","=",slip["employee_id"][0]],
      ["check_in",">=", slip["date_from"]+" 00:00:00"],
      ["check_in","<=", slip["date_to"]+" 23:59:59"],
      ["check_out","!=",false]],
     ["check_in","worked_hours"])
# If wd has a code=="ATTN" line: number_of_hours == sum(att.worked_hours) and
#   number_of_days == count(distinct check_in dates). That's the bridge working.
# If wd has NO "ATTN" line (only calendar codes like WORK100): the employee had NO
#   closed attendance in the window, so the stock calendar fallback ran. That is
#   EXPECTED for salaried/un-clocked employees — not a bug.

5. Refund (credit-note) a confirmed payslip

# Creates a negative "credit note" copy of a done slip (for corrections):
call("hr.payslip", "refund_sheet", [[slip_id]])
# The new slip has credit_note=True and refunded_id pointing at the original.

Pitfalls (everjust-specific and Odoo-19-specific)

1. Check install state before you do anything. Only headsup has the full stack live; several tenants have payroll but not hr_attendance (bridge dormant → stock calendar worked days), and some have no payroll/HR at all. Recipes here silently no-op or error on the wrong tenant. Verify via ir.module.module / describe_model first.

2. "Contract" = hr.version, not hr.contract. Odoo 19 replaced the contract model. Every payroll contract_id (payslip, worked-days line) and the employee's version_id/current_version_id point at hr.version. Searching hr.contract will fail or return nothing. The employee's active contract is version_id.

3. No structure → empty payslip. compute_sheet multiplies worked days by the salary structure's rules. headsup currently has 0 hr.payroll.structure / 0 hr.salary.rule, so computing there yields a slip with worked-days lines but no money line_ids. Ensure a struct_id with rules exists (on the run and/or employee contract) before expecting pay amounts.

4. The bridge only counts CLOSED attendance. _compute_worked_days requires check_out != False. An employee still clocked in (open span) contributes zero hours to the payslip. If worked hours look short, check for open hr.attendance rows.

5. No attendance → stock calendar fallback, and that's intentional. A missing ATTN line does NOT mean the bridge is broken; it means the employee had no closed attendance in the period, so OCA's calendar-based computation ran (salaried staff on a fixed resource_calendar_id). Don't "fix" this by forcing an ATTN line.

6. hr.attendance ≠ account.analytic.line. Attendance (clock) and analytic timesheets are separate systems on this fork — clocking in creates no analytic line. Payroll reads attendance; Projects/timesheets read analytic lines ([[everjust-projects]]). Reconcile payroll against hr.attendance.worked_hours only.

7. hr.payslip.run.state is draft/close — different axis from the slip. The run is a container; close/draft is just its own batch status. The pay workflow (draft→verify→done→cancel) lives on each hr.payslip. Don't confuse them, and don't look for action_payslip_* methods on the run (there are none) — generate slips via the hr.payslip.employees wizard.

8. line_ids and worked_days_line_ids are read-only / computed. Don't write pay lines by hand; they're rebuilt every compute_sheet (which first unlinks the old lines). To change pay, change inputs, the contract wage, or the structure — then recompute.

9. action_payslip_done recomputes by default. Unless prevent_compute_on_confirm is set, confirming re-runs compute_sheet — so late attendance edits made after the first compute WILL be picked up at confirm time. If you need a frozen snapshot, confirm with the without_compute_sheet context or set prevent_compute_on_confirm.

10. Everything is per-company_id and role-bounded. Payslips, attendance, and employees are scoped to the tenant company, and the MCP acts as a specific Odoo user — a limited role may not see payroll data at all (record rules, not a bug). Confirm the tenant DB and your acting user before mutating pay (see [[everjust-platform]] / [[everjust-agent-mcp]]).

11. everjust_brand_hr_attendance is cosmetic only. The sibling auto-install module just white-labels the attendance kiosk "Powered by" footer — it changes no data, no fields, no computation. Never treat it as part of the payroll logic.