CLAUDE FINANCE AGENTS

In May 2026 Anthropic released ten pre-built finance agents in a public GitHub repo. They're the same reference architectures Goldman-tier firms deploy, packaged as free plugins that run in Claude Code and Claude Cowork on any paid plan. Most people have no idea they exist. This guide covers what they are, how to install them, and the prompts I'd run first.

WHAT THEY ARE

Each agent is a template that bundles three parts:

• Skills: the domain playbook. How to build a comp table, what a clean DCF looks like, what belongs in an earnings note.
• Connectors: governed pipes into real data: FactSet, S&P Capital IQ, Morningstar, PitchBook, Daloopa, Moody's, LSEG. Optional; the skills work without them.
• Subagents: extra Claude instances the main agent spins up for sub-tasks like picking comparables or checking methodology.

The repo: github.com/anthropics/financial-services

THE 10 AGENTS

RESEARCH & COVERAGE

• Market Researcher: tracks a sector or theme, synthesizes news, filings, and broker research into an industry view.
• Earnings Reviewer: reads transcripts and filings, updates models, flags what changed versus your thesis.
• Model Builder: builds DCF, LBO, and 3-statement models in Excel from filings and data feeds.
• Pitch Agent: comps, precedents, and LBO analysis assembled into a pitch deck.
• Meeting Prep Agent: briefing packs on any company or counterparty before a call.

FINANCE OPS

• Valuation Reviewer: checks valuations against comparables and methodology standards.
• GL Reconciler: finds breaks in the general ledger and traces root cause.
• Month-End Closer: accruals, roll-forwards, variance commentary.
• Statement Auditor: reviews statements for consistency and audit-readiness.
• KYC Screener: parses entity documents and runs a compliance rules engine.

INSTALL IN 2 MINUTES

# Claude Code: add the marketplace
claude plugin marketplace add anthropics/financial-services

# Core skills first (comps, DCF, LBO, connectors)
claude plugin install financial-analysis@claude-for-financial-services

# Then the verticals you actually need
claude plugin install equity-research@claude-for-financial-services
claude plugin install wealth-management@claude-for-financial-services

In Claude Cowork: Settings, then Plugins, then Add plugin, and paste the repo URL. Pick agents from the marketplace list.

THE COMMANDS YOU GET

• /comps: comparable company analysis
• /dcf: discounted cash flow valuation
• /lbo: leveraged buyout model
• /3-statement-model: full 3-statement build
• /earnings: earnings update note
• /initiate: initiation-of-coverage report
• /screen-deal + /ic-memo: deal screen and investment committee memo
• /client-review, /rebalance, /tlh: portfolio review, rebalancing, tax-loss harvesting
• /debug-model: audits an existing Excel model for errors

PROMPTS WORTH RUNNING

Copy these as written, swap the ticker. Research output, not financial advice; the decisions stay yours.

/comps for [TICKER]. Pull the 8 closest public peers, build the
table on EV/EBITDA, EV/Revenue, and P/E, and flag where it trades
rich or cheap versus the median and why.

/dcf on [TICKER]. State every assumption (growth, margins, WACC,
terminal) in a table before the output, then run bear, base, and
bull cases with the implied share price for each.

/earnings. Here is the latest transcript and 10-Q for [TICKER]:
[paste or attach]. Summarize what changed versus last quarter,
what management dodged, and the three numbers that matter most.

Act as the Market Researcher agent. Map the [SECTOR] landscape:
top players by market share, the two structural tailwinds, the
two biggest risks, and which company screens best on both growth
and valuation.

/initiate coverage on [TICKER]. I want the business model in
plain language, unit economics, competitive moat rated 1-5 with
justification, and the bull and bear thesis in three bullets each.

/rebalance. Here are my current holdings and target allocation:
[paste]. Show the drift per position, the trades that restore
target, and any wash-sale conflicts with the last 30 days.

Start with /comps on a company you already own. You know the answer it should give; if it holds up there, trust it on the ones you don't.

The play: install financial-analysis plus one vertical tonight, run two of the prompts above on a real ticker, and compare the output against your own work. That comparison tells you exactly how much of your research stack this replaces.