Enterprise Business MCP

Multi-domain executive analytics façade · PEF is pack #1 · chat host is separate UI

Service is up · you are viewing this via http://46.250.229.188:8201

How to use (this host)

Hub packs: ["pef", "ncapp", "prs"] · registered systems: ["pef", "odoo", "prs", "ncapp", "context"]

MCP tools

What you can ask (governed metrics)

21 metrics · exploratory until definitions are approved · ask with metric_id or “scenario 1”…“scenario 5”

metric_idTitleDomainSystems ScenarioSourceDescription
active_users_by_departmentActive PEF users by departmentpefpefbuiltinCount of active PEF users grouped by PEF department.
instances_by_statePEF instances by workflow statepefpefbuiltinCount of non-deleted PEF instances grouped by state.
incomplete_pefsIncomplete PEF countpefpefbuiltinPEF instances not submitted and not closed (exploratory incomplete definition).
incomplete_by_departmentIncomplete PEFs by departmentpefpefbuiltinIncomplete PEF instances joined to employee PEF department.
instances_by_cyclePEF instances by cyclepefpefbuiltinCount of PEF instances per evaluation cycle.
workflow_status_by_company_deptWorkflow status by company & department (published cycles)pefpef1builtinFor every PUBLISHED evaluation cycle: PEF counts by workflow state, grouped by company and department, with group totals, state share %, completion % (CLOSED),
eligibility_vs_assignmentEligibility vs assignment (regularization cutoff)pefpef2builtinFor each PUBLISHED cycle: (a) active REGULAR employees regularized on/before cutoff without a PEF assignment; (b) assigned employees regularized after cutoff.
overdue_pef_activitiesOverdue PEF activities (milestone-based)pefpef3builtinAs of today, open PEFs past the cycle milestone mapped to their current state (self-rating, KPM, calibration, close). Shows employee, manager, department, activ
self_vs_manager_disagreementTop self vs manager rating disagreementspefpef4builtinFive largest disagreements between SELF and MANAGER ratings on matching People Analyzer, GWC, or KPM rows (same row_id).
gwc_wrong_seat_kpm_passed_auditAudit: GWC WRONG_SEAT with KPM PASSEDpefpef5builtinPEFs where GWC is WRONG_SEAT and KPM is PASSED. Recalculates KPM as weighted average of manager ratings (pass >= 3.00) and next step from PA×GWC matrix; flags d
results_recalc_mismatchesResults recalculation mismatches (broader audit)pefpef5bbuiltinBroader audit: PEFs whose stored KPM score/result or next_step differs from recalculation (manager KPM weighted avg; PA×GWC next-step matrix).
ncapp_productions_by_monthNCAPP productions by monthncappncappbuiltinCount of Companion App loan productions grouped by created month.
ncapp_productions_by_branch_codeNCAPP productions by branch codencappncappbuiltinProductions grouped by product_number prefix (3-letter branch code). Code is not an official org branch id until mapped in context.
ncapp_productions_by_sourceNCAPP productions by sourcencappncappbuiltinProductions grouped by source_of_production (Branch Release, Telemarketer, etc.).
ncapp_par_current_by_tagNCAPP current PAR tagsncappncappbuiltinDistribution of current portfolio-at-risk (PAR) tags on accounts.
ncapp_payment_posting_statusNCAPP payment posting by category/statusncappncappbuiltinPayment posting rows by category and execute_status (top combinations).
prs_parcels_by_statusPRS parcels by statusprsprsbuiltinOutbound parcels (not soft-deleted) grouped by status.
prs_parcels_by_monthPRS parcels by monthprsprsbuiltinOutbound parcels by created month.
prs_parcels_by_branchPRS parcels by branchprsprsbuiltinOutbound parcels by branch (transmitter id joined to branch).
prs_parcels_by_courierPRS parcels by courierprsprsbuiltinOutbound parcels grouped by courier_name.
learned_pending_kpm_bae721d966Pending KPM count (learned proof bae721d966)pefpeflearnedlearnedLearned metric from teaching session. Gap: new_metric Systems: pef Domain: pef Intent: Count PEFs stuck in KPM step for ops follow-up Population: Non-deleted PE
If the AI cannot answer

It must refuse (hard stop), then proactively ask in the same turn if you want to teach a new metric — including for unknown abbreviations. It must not open application codebases or invent SQL. Learning is multi-domain (systems first) — not PEF-only forever.