Kalc
Request beta access
Private beta — invitation only

Meet your Independent Financial Audit Analyst.

Audit capacity for every team — an independent verification pass on structure, calculations, and financial logic, and whether the statements and the saved case hold together. Clear, severity-ranked findings in minutes, before the lender call, IC, or external review.

Request beta access See how it works →
addin.kalc.tech — Kalc Illustrative
Kalc Project Finance model · illustrative
ASSURANCE SCORE
81High confidence
2,847 formulas 47 findings 3 critical

Three covenant and structure issues need attention before external review.

  • DSCR breach in years 3–5
  • Hard-coded discount rate
  • IRR construction vs sector norm
Export analysis report ↗

Illustrative example · same three tabs as the live Excel add-in

How it works

Three kinds of intelligence. One clearer picture.

Kalc applies systematic checks first, AI-assisted insights second — so every finding is traceable to a cell, not a guess.

A thorough external review follows a defined process — scope, iterations, and a professional conclusion. Kalc runs the systematic verification pass inside that workflow: before the pack leaves, between cycles, and after every material change.

∑

Systematic checks

Every formula reconstructed from source. Every dependency mapped. Structural clarity before AI adds context.

◎

AI-assisted insights

Sector-aware logic checks on verified data — DSCR, IRR, NPV, and covenant patterns — explained with cell-level references.

✓

Actionable output

Assurance score, severity-ranked findings, and an exportable analysis report — ready for your working papers and presentations.

The analysis flow

From raw model to informed next step

Understand

Structure mapped. Dependencies traced.

Validate

Formulas checked. References flagged.

Interpret

Logic assessed for purpose.

Prioritise

Findings ranked by severity.

Share

Exportable report for your review.

The problem

Model completion is not decision confidence

◇

LLMs produce probable answers

Financial models are computation graphs, not text. A language model describes what a formula likely does — it cannot reconstruct every path and prove correctness.

◷

Specialist review is hard to scale

The thorough model review you want before sharing a workbook is slow, expensive, and inconsistently documented — especially when timelines are tight.

◈

Evidence gets buried

Findings scattered across chat transcripts, email threads, and verbal sign-off. Teams need cell-level references, severity scores, and a reproducible analysis record.

What you get

Four questions. One analysis.

Structure, statements, the saved case, and the evidence trail — checked in the order a reviewer actually asks them.

Why verification capacity is the constraint in 2026 →

Does the model hold together?

Structural integrity

Every formula reconstructed. Every dependency traced. Defects ranked, not dumped. See what a change touches before you make it.

Does the finance hold together?

Financial assurance

Where the model presents statements, Kalc checks whether they tie, roll, and hold together — including the model’s own check cells.

What does it actually say?

Scenario & sensitivity

See which case the workbook was saved on — and which drivers each case actually moves — before covenants are read on the wrong case.

What evidence does the analysis leave?

Evidence & record

Documentation coverage, version comparison, and an exportable analysis record. Not a sign-off. You remain responsible for the analysis.

Same workflow in Excel: severity-ranked findings, an exportable report, and — for agent stacks — findings via API or MCP. Kalc verifies; it never builds.

Built for your work

Verification that fits how you already work

Guides for modellers and compliance reviewers — and a workflow that extends formal review, not replaces it.

For modellers →

First verification pass, assumptions, Zero Retention, and when to escalate to human review.

For compliance reviewers →

Data handling, Contained vs Grounded, GDPR rights, and a review checklist.

Practitioner guides →

Assumptions documentation, pre-decision verification, and how Kalc fits your workflow.

Extends your review workflow

Read the guide →

Before external review

Fix clear defects and arrive with a severity-ranked findings summary for your reviewer.

During iteration

Re-run after material edits; see what resolved and what regressed between versions.

Between engagements

Maintain baseline model hygiene when a full review is not yet booked.

Complements build tools

Build and verify →

Copilots draft

Verify reconstruction and consistency after drafting — in the same workbook.

Agents populate

Receive agent-ready findings; fix and re-verify in the loop via API or MCP.

You decide

Kalc verifies; it never builds. Findings identify and suggest — you choose what to fix.

For AI-built models

The agent builds. Kalc verifies.

AI can build a model in minutes — but it cannot independently check its own work. Kalc is the separate verification step: deterministic findings your agent can act on via API or MCP.

⟳

Verify in the loop

Call Kalc from your agent over API or MCP. Every formula, dependency, and financial-logic rule checked deterministically — not guessed. Findings anchored to the exact cell.

✓

Findings your agent can act on

A remediation brief for your build agent: what to change, where, and why — grouped by severity. Re-submit and see what resolved or regressed.

Built for agent workflows
  • POST /sessions — submit a version, get findings, deltas, and a gate status your orchestrator can branch on.
  • GET /audit/{id}/brief.md — an agent-ready remediation brief for your build agent.

Available over REST and MCP. Kalc verifies; it never builds. Findings identify and suggest — not legal, financial, or professional advice.

Request beta access
Built for trust

Confidential by architecture — not by policy

EU-hosted processing

Model analysis runs on EU infrastructure. No transatlantic transfers of your deal data are described in our published Privacy Notice.

Analysed in memory. Never stored.

Your model is analysed in memory and never written to our storage. Nothing is written to disk, nothing is backed up, nothing is kept beyond the run — and no AI provider we use trains on it.

GDPR Article 28

Data processing agreement available. You remain the controller; Kalc acts as your processor.

Separate from authorship

Kalc did not build your model and cannot edit it. It identifies issues and suggests remediation — analytical support only, not legal, financial, or professional advice. Kalc verifies and recommends; you remain responsible for the analysis.

Beta program

Shape the product from day one

✦

Full analysis access during beta

Full analysis at no cost during the invitation-only learning cohort. Project Finance, M&A, Infrastructure, Real Estate, and VC vertical templates supported.

✉

Direct line to the team

Invited beta testers join a private Slack channel with the people building Kalc — plus chat and email. Not a support queue.

◎

Influence the roadmap

We are looking for analysts who will run Kalc on real models and tell us, candidly, what they find within fourteen days.

⬡

Early-adopter pricing at launch

Beta participants receive locked early-adopter pricing when Kalc launches commercially.

addin.kalc.tech — Analysis Overview
ASSURANCE SCORE
81High confidence
Debt!B12 DSCR covenant breach Minimum 1.25× violated in years 3–5 · 96% confidence
Inputs!C42 Hard-coded discount rate 8.5% embedded in formula chain · 94% confidence
Returns!D18 Non-standard IRR construction Sector convention deviation · 87% confidence
Export analysis report ↗
Get early access

Apply for beta access

We are inviting a small number of independent consultants for this learning cohort. Tell us about your work and we will respond within 48 hours.

  • Full analysis at no cost during beta
  • Direct access to the Kalc team
  • Early-adopter pricing locked at commercial launch

By applying you agree to our Privacy Notice and Beta Terms. We never share your details with third parties.

Questions

Frequently asked

What does Kalc analyse? +

Excel financial models — whether the structure holds, whether the statements hold where they are presented, which case the workbook was saved on, and what evidence the analysis leaves. Severity-ranked findings with cell-level evidence, directly inside Excel.

How is this different from ChatGPT or Copilot? +

Language models produce probable descriptions of formulas. Kalc applies systematic checks first, then AI-assisted insights — with cell-level references you can review. Kalc did not build your model.

How does Kalc fit with external model review? +

Kalc is designed as a preparation and follow-up layer around the review you already use — run passes between meetings, fix clear defects early, and arrive with a findings summary. It does not sign off models or replace a qualified external reviewer. Read how they work together →

Where is my data processed? +

On EU infrastructure. Your model is analysed in memory for the duration of the run. You can use Zero Retention so nothing from the analysis is stored after it finishes. New workspaces default to Contained — nothing about your model is sent to a third party unless you opt in to Grounded advisory features. See our Privacy Notice.

What is included during beta? +

Full access at no cost for 30 days from your first analysis, extendable on material contribution — see our Beta Terms. Direct support from the Kalc team. We request first impressions within fourteen days of your first successful run.

Which sectors are supported? +

Project Finance, M&A, Infrastructure, Real Estate, and VC templates are in active beta. General financial logic checks apply to all models regardless of sector.

Every model deserves a review.

Kalc gives your team the audit capacity to get there — in minutes, inside Excel.

Request beta access →