Ask your financial data anything.
Get answers in seconds.
Your finance team's answers live in Essbase, Planning, and FCCS. AgentCube connects Claude, Copilot, ChatGPT, or Gemini to the cubes you already trust, so anyone can query them in plain language.
Why did Germany miss Q3 EBITDA plan by $4.2M?
Pulling Q3 actuals and the plan scenario for Germany.
planning.query_data(entity="DE", scenario="Plan_Q3_FY26")essbase.mdx(cube="Actuals", filter="DE_Q3")Germany came in $4.2M below plan on Q3 EBITDA. Three drivers account for $3.9M of the gap.
| Driver | Entity | Variance |
|---|---|---|
| Raw material COGS | DE-Manufacturing | −$2.1M |
| FX (EUR/USD) | DE-Consolidated | −$1.1M |
| Channel mix, retail | DE-Sales | −$0.7M |
| All other | DE (residual) | −$0.3M |
Want me to drill into raw material COGS by supplier, or pull last year's Q3 for context?
Where does the October close stand? Any entities still open?
Checking Financial Consolidation and Close for period FY26-Oct.
fcc.get_close_status(period="FY26-10")October close is eighteen of twenty-two entities complete. Four still open:
| Entity | Status | Blocker |
|---|---|---|
| DE-Sales | Pending review | Three unposted journals |
| BR-Consolidated | Rejected | Reconciliation variance $180K |
| UK-Holdings | In progress | Intercompany elim, pending |
| APAC-Elim | Not started | Awaiting JP close |
Want me to message the preparers a reminder, or pull the Brazil reconciliation detail?
Build me a 2027 Germany forecast that assumes we hold FY26 drivers flat except volume grows 4 percent.
Loading Germany FY26 driver set from Planning and applying the four-percent volume assumption.
planning.get_form(form="DE_Driver_Forecast")planning.project_scenario(base="FY26", volume="+4%")Draft FY27 Germany forecast, FY26 drivers held, volume +4 percent:
| Metric | FY26 | FY27 (draft) |
|---|---|---|
| Revenue | $142.0M | $147.7M |
| Gross margin % | 38.2% | 38.2% |
| Operating expense | $31.4M | $32.7M |
| EBITDA | $22.8M | $23.7M |
Draft saved to a working scenario. Want me to route it to your FP&A lead for refinement, or test a second volume sensitivity?
AgentCube retrieves numbers from the systems your finance team already operates, with the user's identity and permissions intact. Examples are illustrative; actual responses reflect your data.
How AgentCube knows what you mean.
Your team asks in the words they already use. AgentCube resolves each term against the way your company actually operates — your close calendar, your ledger, your planning system — and what it doesn’t know yet, it proposes for a person to approve. Same definitions, same answer, whichever AI your team works in.
Controller“Where did Q1 revenue land against plan for the core segment?”
FP&A analyst“How did first-quarter net revenue track to budget in our core BUs?”
Regional finance lead“Q1 rev vs plan, core segment?”
CFO“Did we hit plan on core-segment revenue in Q1?”
- Period
- Apr to Jun
- your close calendar
- Account
- Net revenue
- your ledger
- Version
- Approved working plan
- your planning system
- Entity
- Three business units
- learned from your team, approved by your admin
Want the technical details?Read the full architecture
Designed around the controls your security team already enforces.
AgentCube is a thin layer over the systems you already run. It respects the identity model, the role hierarchy, and the audit posture your team has already invested in.
Identity-first
Every query runs as the requesting user, against the same roles they have today. The model never sees data the analyst could not already see at their desk.
Isolation
AgentCube is not a hosted service. You deploy the stack inside your own network, with no shared Caprus backend in the middle and no other tenants to leak across.
Audit trail
Every tool call logs the user, data source, and request shape, never the response. The hash chain makes tampering evident, so compliance can audit without parsing chat transcripts.
No data retention
Result payloads pass through AgentCube to the model and are never persisted. The data lives in your source system, before and after the query. AgentCube does not keep its own copy.
Deploy where your data lives.
AgentCube ships as container images you run inside your own cloud or on-premise environment, on whatever infrastructure your security team already trusts. No data ever leaves your perimeter to reach AgentCube.
Your stack, not ours.
AgentCube doesn't ask you to move your data, replace a system, or standardize on one AI. These are the systems and models our team works in today, and the list grows with the engagements.
Financial Systems
Data Platforms
AI Platforms
The questions every CFO asks before buying.
Ready to Transform Your Finance Operations?
Connect with our experts today and discover how CAPRUS can unlock your finance team's full potential.
Connect