Know your runway 13 weeks out

Founder with no CFO guessing runway in a sheet · Assemble in ~20 min; first complete forecast in 30-60 min depending on how messy the inputs are.

FinanceEdit flowon X◆ new (unverified) 96

step 1partial

Pull live bank balance and recent transactions

cashflow

step 2ready

Map actuals — categorize every dollar in and out

finance4 alts

step 3ready

Compute burn rate and current runway

ops-revenue

step 4ready

Run actuals retro — spot anomalies vs. prior period

money-retro3 alts

step 5ready

Build 13-week cash-flow table with scenario toggles

financial-model-architect4 alts

step 6ready

Variance analysis — week-over-week rolling diff

financial-health

step 7ready

CFO stress-test and cash management recommendations

cfo-advisor4 alts

step 8ready

Produce one-page runway report

financial-model-review3 alts

Did this recipe work for you?

What you get

▸One-page runway report (Week-0 = bank balance, burn within $1 of manual, rows sum)
▸13-week cash-flow table with base / upside / downside scenario columns
▸Variance report — actuals vs. prior period, line-by-line delta
▸CFO stress-test flags and prioritized cash-management action list

tools: Plaid developer account (free sandbox; production from $0.30/connected account/mo) · No other paid tools required

tokens: ~$2-5 per weekly run at typical startup complexity

≈ $5-15/mo all-in at weekly cadence, dominated by token cost not tools

What stays yours

· Accountant or bookkeeper review before sharing the report with investors or lenders — AI-assembled forecasts are useful internal tools but material financial communications warrant professional sign-off
· Validate category assignments in the actuals map for any novel transaction types (one-time payments, grants, loan draws)
· Set and update scenario assumptions (revenue growth rate, headcount timing) — the model applies your numbers, it does not predict them

The problem

When I need to know exactly how much runway I have and where cash is going each week, I want one session that pulls real bank balances, maps every inflow and outflow to named sources, builds a 13-week rolling table, and runs base/upside/downside scenarios — so I can show investors and my own gut a single defensible number.

Cybernetic layers

Memorybyo

Prior-period actuals must be saved manually (CSV or Google Sheet) so the retro step has something to diff against. No harvested memory artifact covers rolling-period financial state yet.

Background-schedulingpartial

Claude Code cron can re-run this recipe weekly. However, the Plaid bank pull (step 1) requires an interactive auth token refresh when the session expires — fully unattended weekly runs depend on token persistence, which Plaid's sandbox supports but production requires a refresh flow.

Harnessready

Claude Code is the runtime. Drop each skill into ~/.claude/skills/ and invoke by name. The cashflow plugin (step 1) installs via its plugin.json.

Click any step to see what fills it — and the ranked alternatives you can swap in.