Financial Analysis: The Core Plugin
In Lesson 4, you explored the knowledge-work-plugins/finance plugin -- the corporate finance plugin built for controllers and FP&A teams. That plugin serves operational finance: keeping the books clean and the close on schedule. Now you move to a different audience entirely.
The anthropics/financial-services-plugins suite serves investment professionals -- analysts, bankers, portfolio managers, and researchers whose work centres on valuation, deal execution, and market analysis. This lesson covers the foundation of that suite -- the financial-analysis core plugin that every add-on depends on. You will install it, run commands hands-on, and interpret the output -- the same pattern as Lesson 4, but for investment analysis rather than corporate accounting.
Install the Core Plugin
Every plugin in the financial-services-plugins suite shares one architectural decision: all data connectors live in the core plugin. Install the core once, configure your data connections once, and every add-on you install afterwards inherits them automatically. Install order is mandatory -- core first, then add-ons.
- In the Cowork sidebar, click Customize → Browse plugins → Personal.
- Click the + button, then select Add marketplace from GitHub.
- Enter the repository URL:
https://github.com/anthropics/financial-services-plugins - The marketplace loads with all available plugins. Find Financial Analysis and click Install.
Once installed, the plugin activates automatically. Skills are applied when relevant, and slash commands become available in your Cowork session.
Verify the install by typing /comps in the Cowork chat. You should see the command auto-complete. If it does not appear, return to Customize → Browse plugins and confirm the plugin shows as installed.
Data Connectors
The core plugin can connect to eleven data providers. Each requires a separate subscription or API key from the provider. You configure credentials once in the core plugin; every add-on uses them automatically.
| Provider | What It Provides |
|---|---|
| Daloopa | Automated financial model data extraction |
| Morningstar | Equity research, fund analysis, valuation data |
| S&P Global | Capital IQ financial data, credit ratings, ownership |
| FactSet | Market data, financial analytics, portfolio data |
| Moody's | Credit ratings, fixed income analytics |
| MT Newswires | Real-time financial news feeds |
| Aiera | Earnings call transcripts, corporate event intelligence |
| LSEG | Market data, analytics |
| PitchBook | Private company, VC, and PE deal data |
| Chronograph | Private markets portfolio monitoring |
| Egnyte | Document management for deal files |
You do not need all eleven connectors to use the plugin. Without connected data, you provide the figures manually. With connected data, commands pull automatically.
WACC -- Weighted Average Cost of Capital -- is the discount rate used in a DCF. It represents the blended required return of both debt and equity investors, weighted by their proportions in the capital structure.
Formula: WACC = (E/V x Cost of Equity) + (D/V x Cost of Debt x (1 - Tax Rate))
Where E = market value of equity, D = market value of debt, V = E + D. The tax adjustment reflects that interest on debt is tax-deductible, reducing the after-tax cost of debt.
Cost of equity is estimated via CAPM: Cost of Equity = Risk-Free Rate + Beta x Equity Risk Premium. The risk-free rate is the 10-year Treasury yield. Beta measures the company's stock volatility relative to the market.
Why WACC matters: A 1 percentage point increase in WACC typically reduces DCF equity value by 10-20%. This sensitivity is why the WACC/terminal growth rate sensitivity table is always included in professional analysis.
Example: A company has 60% equity, 40% debt. Cost of equity: 12%. Cost of debt: 5%. Tax rate: 25%. WACC = (0.60 x 12%) + (0.40 x 5% x 0.75) = 7.2% + 1.5% = 8.7%.
Enterprise value is what you would need to pay to acquire a company entirely, including both equity and debt.
Formula: EV = Market Cap + Total Debt + Preferred Stock + Minority Interest - Cash
Why subtract cash? You inherit the company's cash when you acquire it, which offsets part of the price. A company with $500M market cap, $200M debt, and $50M cash has an EV of $650M.
Example: Company A has 100M shares at $8.50. Market cap = $850M. Debt = $150M. Cash = $30M. EV = $850M + $150M - $30M = $970M. With LTM EBITDA of $97M: EV/EBITDA = 10.0x.
Why use EV over market cap? EV is capital-structure-neutral. Two companies with identical operations but different debt levels have different market caps but similar EVs. EV/EBITDA allows fair comparison between differently leveraged businesses.
LTM (Last Twelve Months) refers to a company's financial performance over the most recent twelve-month period, regardless of fiscal year end. Used in comps to ensure all companies are compared on the same timeframe.
Calculation: LTM = Most Recent Annual Period + Most Recent Partial Period - Prior Year Comparable Partial Period.
Example: A company's fiscal year ends December 31. It is now October. LTM revenue = FY2024 revenue + Q1-Q3 2025 revenue - Q1-Q3 2024 revenue. This gives the twelve months ending September 30, 2025 -- the most current full year available.
A Leveraged Buyout (LBO) model values the acquisition of a company using significant borrowed money -- typically 50-70% debt. The acquired company's own cash flows service the debt. Private equity firms use LBOs to acquire companies, improve them over 3-7 years, and sell at a profit.
Why debt amplifies returns: Buy a company for $100M using $30M equity and $70M debt. Sell for $130M five years later after repaying debt from operating cash flows, leaving $60M net. Equity return: $60M on $30M invested = 100% total return (15% IRR). Without leverage, the same investment yields 30% total. Debt amplification works in both directions -- underperformance can wipe out equity entirely.
Key metrics: Entry multiple (EV/EBITDA at purchase), exit multiple (EV/EBITDA at sale), IRR (annualised equity return, target 20-25%), MOIC (total proceeds / equity invested, target 3-5x over 5 years).
The Eight Core Commands
| Command | What It Produces |
|---|---|
/comps [company] | Comparable company analysis as Excel workbook -- peer group, multiples, implied valuation range |
/dcf [company] | DCF valuation model with sensitivity tables (WACC x terminal growth rate) |
/lbo [company] | LBO model with entry/exit scenarios, debt schedule, and returns summary |
/3-statements [company] | Three-statement financial model (income statement, balance sheet, cash flow) |
/competitive-analysis [company] | Competitive landscape analysis -- market positioning, share, and strategic context |
/debug-model | Debug and validate a financial model for formula errors, circular references, and logic gaps |
/check-deck | Review a pitch deck for numerical consistency, formatting, and narrative coherence |
/ppt-template | Register the firm's branded PowerPoint template for all deck outputs |
Nine passive skills complement these commands: 3-statements, check-deck, check-model, competitive-analysis, comps-analysis, dcf-model, lbo-model, ppt-template-creator, and skill-creator. When you run /comps, the comps-analysis skill is also active in the background -- applying professional conventions beyond what the command explicitly covers. When you paste a financial model without invoking any command, the check-model skill activates and flags potential issues automatically.
Worked Example: Comps Analysis
You are an equity analyst who needs a quick market check on a mid-cap industrial company before a client meeting tomorrow morning. You type:
/comps Acme Industrial
Then provide context: "My subject company has LTM Revenue of $340M and LTM EBITDA of $51M."
Claude queries the configured connectors -- FactSet for market data, Daloopa for standardised financial model data -- pulls comparable public companies in the industrials sector, calculates enterprise values, and builds the workbook.
The output is a formatted Excel workbook following professional conventions: blue input cells for manually adjustable figures, black formula cells, green output cells for the multiples table. The peer group table shows each company's LTM Revenue, LTM EBITDA, EBITDA Margin %, EV, EV/EBITDA, and EV/Revenue. Summary statistics rows show Median, 25th Percentile, and 75th Percentile.
You follow up: "Based on this peer group, what is the implied enterprise value range for my company?"
Claude responds with the analysis: your EBITDA margin of 15.0% is below the peer median of 18.3%. Applied to your EBITDA of $51M at the 25th percentile (8.1x), implied EV is $413M; at median (9.2x), $469M; at 75th percentile (11.4x), $581M. A buyer applying a discount for the margin gap may use 7-8x, implying $357M-$408M.
The professional skill is not the arithmetic. The skill is evaluating whether the peer group is right, whether the margin gap is structural or correctable, and what that implies for the appropriate multiple.
Exercise 8: Comps and DCF
Time: 30 minutes. Requires: Cowork with the financial-analysis core plugin installed.
-
Run
/compsfor any publicly listed company you follow. If you have connected data providers, the command pulls financials automatically. If not, provide the company's LTM Revenue, LTM EBITDA, and sector. Review the Excel workbook output: check the peer group, the multiples table, and the implied valuation range. -
Run
/dcffor the same company. Review the sensitivity table (WACC vs terminal growth rate). Locate the current share price in the grid -- which combination of WACC and terminal growth rate justifies the market price? Is that combination reasonable? -
Compare the two valuations. Where do the comps-implied EV range and the DCF-implied equity value overlap? Where do they diverge? Divergence tells you about different assumptions -- comps reflect market sentiment while DCF reflects your projected cash flows.
-
Ask Claude: "Draft a one-paragraph valuation summary for an investment committee. Reference both the comps range and the DCF base case, and explain the key driver of any gap between them."
The discipline: Two methodologies should converge. When they diverge significantly, the gap is telling you something about the market's assumptions versus yours. The professional skill is explaining that gap, not ignoring it.
Exercise 9: Debug and Check
Time: 15 minutes. Requires: Cowork with the financial-analysis core plugin installed.
-
Run
/debug-modelon a financial model you have built or inherited. If you do not have one handy, ask Claude: "Build a simple three-year revenue projection model with a circular reference between revenue growth and operating expenses. Then run /debug-model on it." Review the error report: does it correctly identify the circular reference? -
Run
/check-deckon a presentation draft (PowerPoint or Google Slides). If you do not have one available, ask Claude: "Create a two-slide earnings summary with a deliberate inconsistency: the revenue figure on slide 1 should differ from the revenue figure in the chart on slide 2. Then run /check-deck." Review the consistency report.
The discipline: Verification commands are not optional -- they are the production discipline. /debug-model catches formula errors before they reach a client. /check-deck catches number inconsistencies before they reach a committee. Building the habit of running verification after every model change is what separates reliable output from risky output.
Try With AI
Use these prompts in Cowork or your preferred AI assistant to explore the core plugin commands.
Prompt 1: Plugin Command Selection
I work in [YOUR FINANCE FUNCTION -- e.g., equity research,
investment banking, corporate development, FP&A].
My most common analytical task is [DESCRIBE THE TASK --
e.g., "valuing acquisition targets", "preparing board
presentations with peer benchmarks", "building projection
models for budget season"].
Based on the eight core plugin commands (/comps, /dcf, /lbo,
/3-statements, /competitive-analysis, /debug-model,
/check-deck, /ppt-template):
1. Which command maps to my task?
2. What would the output look like?
3. Which data providers would supply the underlying data?
4. What would I still need to review manually before the
deliverable is ready for my audience?
What you're learning: Mapping real workflows to plugin commands requires understanding what each command produces and which data sources feed it. This prompt practises the three-layer mapping -- command, data provider, manual review -- that separates someone who installs a plugin from someone who integrates it into professional practice.
Prompt 2: Sensitivity Table Interpretation
A DCF model for Company X produces this sensitivity table
showing equity value per share at different WACC and terminal
growth rate combinations:
Growth 1.5% Growth 2.0% Growth 2.5%
WACC 8.0% $42.10 $47.30 $54.20
WACC 9.0% $33.80 $37.10 $41.50
WACC 10.0% $27.60 $29.80 $32.70
The current share price is $35.00.
1. What combination of WACC and growth rate justifies the
current market price?
2. Is that combination reasonable for a mid-cap industrial
company? Why or why not?
3. What is the spread between the highest and lowest values
in this table? What does that spread tell a decision-maker
about how much confidence to place in any single DCF number?
4. If you were presenting this to an investment committee,
how would you frame the uncertainty?
What you're learning: A sensitivity table is the most honest output of a DCF because it shows the range of outcomes rather than pretending one number is correct. Reading a sensitivity table is a professional skill: you need to locate where the current market price sits in the grid, assess whether those assumptions are plausible, and communicate the range to decision-makers who want a single answer.
Prompt 3: Chaining Commands
A corporate development team is evaluating whether to
acquire a competitor. Map the full analysis sequence:
1. Run /comps to establish the market context -- what do
peers trade at?
2. Run /dcf to build a fundamental valuation -- what is the
target worth based on projected cash flows?
3. Run /lbo to test sponsor returns -- what could a
financial buyer achieve?
For each step:
- What output from the previous step informs this one?
- What assumption must I validate before proceeding?
- What professional judgment cannot be automated?
Then: if comps imply 8-10x EV/EBITDA but the DCF implies
12x, what does that divergence mean and how would you
present it to the board?
What you're learning: Individual commands solve individual problems; professional workflows chain commands together. The order matters -- comps establish the market context that informs DCF assumptions, which in turn anchor the LBO entry multiple. Understanding this chain reveals how the core plugin supports an end-to-end analysis rather than isolated calculations.
Flashcards Study Aid
Continue to Lesson 6: Domain Plugins -- From Deals to Portfolios →