MCP plugin — Claude Code · Claude Desktop · Codex

Give Claude a finance workspace.

You use Claude on spreadsheets. You know the limits — it drifts on formulas, forgets between sessions, burns tokens on grids. Layerz holds the structure so Claude doesn't have to.

Your data stays yours · Use your own Claude · Export to .xlsx

Bump churn to 4% on DistribSaaS. What changes for Q1?

I ran it as a dry-run, so the saved model is untouched. Here's the Q1 impact:

Layerz patch
Net revenue
€428K
−3.0% from €441K
Gross margin
€176K
−7.1% from €189K
Runway
13.4 mo
−1.6 from 15 mo

Save as scenario_churn_4% and stress a downside?

Yes, and add a downside at −15% revenue

Scenario saved. The downside cuts deeper:

Layerz patch
EBITDA Y1
€212K
−51.9% from €441K
Runway
8 months
−5.2 from 13.4 mo
Covenant DSCR
0.88×
breach

Covenant breaks at month 9. Want me to surface a debt restructuring option, or just flag it on the board pack?

Plot quarterly EBITDA, base vs downside

Quarterly EBITDA, both scenarios:

Layerz read
Base caseDownside −15%EBITDA · €k
Covenant floor
Q1Q2Q3Q4

The downside sits under the covenant floor from Q1 onward.

Flag it. Then export both scenarios for the board.

Done. Board pack ready:

Layerz export
DistribSaaS_Q1_board.xlsx
Export Layerz · 92 KB · 2 sheets

Standard Excel, no Layerz account needed on their end.

Write a message…
Opus 4.8

Live demo · Layerz answers Claude in real time

Where Claude breaks

Finance pros who hit the same wall.

Claude in Excel is awesome for a first draft. But the moment I shape it around how we actually operate, rows shift and a formula on the home stretch lands on the wrong cell. Not something I’d ship.

Modeler, finance consultancy

Every session I open, I’m re-explaining the same model — the assumptions, the structure, where we left off. I feel like I never finish onboarding it.

CFO, PE-backed SME

A few rounds into a real spreadsheet and the conversation is already too long. I cut it into pieces, lose half the context, and end up tracking what changed in my head.

Associate, transaction advisory

What the model looks like

The missing layer for working safely with AI in finance.

Every item is named. Every formula is traceable. Claude navigates the dependency graph instead of scanning cells.

P&LSaaS · 3 years
AMRR€42K
AChurn rate3.2%
FNet revenue€40.6K
KGross margin68%

This is why your agent stops drifting — the model is read, not guessed. Every dependency holds across sessions.

Your existing workflow

Same deliverables. Your agent does the work.

Monthly close, board pack, cash forecast — unchanged for your colleagues. Standard .xlsx, no Layerz account on their end. Your agent stops drifting on the way there.

P&L · 2026 – 2028live
Y1Y2Y3
Revenue€1.2M€1.9M€2.8M
COGS€0.3M€0.5M€0.6M
Gross margin€0.9M€1.4M€2.2M
Opex€0.7M€1.0M€1.4M
EBITDA€235K€441K€523K
EBITDA bridge · Y1 → Y3+122%
€235k
+340
−52
€523k
Y1 EBITDA+ Revenue growth− Cost growthY3 EBITDA
Things you can ask Claude
How to connect

Pick your agent. Paste. Go.

claude mcp add -t http layerz https://app.layerz.cc/mcp

Run it, then `claude /mcp` to authenticate via OAuth.

How it fits your stack

Built for financial logic, not generic data.

FeatureLayerzExcelClaude alone
Dynamic calculations (formulas, dependencies)✗ raw figures
Structured for AI navigation✗ cells only✗ no persistence
Variables with type + timeline
Persists context across sessions✓ data only
Native financial logic (dependencies, patterns)✗ manual✗ raw figures
Versioning — every edit logged, auditable, revertable✗ file copies
Export to standard .xlsxnative

Keep using Claude. Now with financial framework.

Your data stays yours · Use your own Claude · Export to .xlsx

Layerz | The MCP plugin for financial modeling with Claude