Anthropic Open-Source Financial Services Toolkit Turns Claude into a Wall Street Analyst

Financial Services Toolkit

Anthropic open‑sourced a financial‑services toolkit that extends Claude into a specialist analyst for investment banking, research, private‑equity and wealth‑management workflows. The toolkit provides end‑to‑end pipelines from data acquisition through modeling to document generation.

MCP Direct Connections (11 Data Providers)

S&P Global – Endpoint: https://kfinance.kensho.com/integrations/mcp – Enterprise ratings, financials, industry benchmarks, capital‑structure data for valuation and risk control.

Moody's – Endpoint: https://api.moodys.com/genai-ready-data/m1/mcp – Credit ratings, default probabilities, bond and issuer risk data for credit review.

Daloopa – Endpoint: https://mcp.daloopa.com/server/mcp – Structured financial statements, metric extraction, model validation for financial modeling.

FactSet – Endpoint: https://mcp.factset.com/mcp – Global equity, fixed‑income and commodity market data for market analysis.

LSEG – Endpoint: https://api.analytics.lseg.com/lfa/mcp – Interest curves, FX, bond relative value, volatility for fixed‑income analysis.

MT Newswires – Endpoint: https://vast-mcp.blueskyapi.com/mtnewswires – Real‑time news, announcements, market events for event‑driven research.

Aiera – Endpoint: https://mcp-pub.aiera.com – Earnings call transcripts, event‑driven data, speech‑to‑text for earnings analysis.

Morningstar – Endpoint: https://mcp.morningstar.com/mcp – Fund data, holdings, performance attribution, risk metrics for asset‑management.

PitchBook – Endpoint: https://premium.mcp.pitchbook.com/mcp – PE/VC deal, M&A, project and exit data for due‑diligence.

Chronograph – Endpoint: https://ai.chronograph.pe/mcp – PE/fund portfolio management, post‑investment monitoring, performance reporting.

Egnyte – Endpoint: https://mcp-server.egnyte.com/mcp – Enterprise document storage, due‑diligence files, compliance records.

Financial Agents (10) covering Front‑, Middle‑ and Back‑Office

Research & Client Coverage (Front‑Office)

Pitch builder – Generate target lists, comparable analysis, auto‑create pitchbooks.

Meeting preparer – Auto‑assemble client briefs and meeting materials.

Earnings reviewer – Interpret earnings releases and call transcripts, update models, flag investment‑logic changes.

Model builder – Automatically build and maintain financial models from filings.

Credit Risk & Compliance (Middle‑Office)

Market researcher – Track industries and issuers, tag credit‑risk signals for decision support.

KYC screener – Collect due‑diligence documents, audit source files, package compliance deliverables.

Finance & Operations (Back‑Office)

Valuation reviewer – Valuation cross‑check, comparable benchmarking, methodology validation.

General ledger reconciler – GL reconciliation, NAV calculation, variance tracing.

Month‑end closer – Automate month‑end voucher generation and closing reports.

Statement auditor – Check financial‑statement consistency, completeness, auditability.

Vertical Plugins (7) forming standardized workflows

financial-analysis (core) – Comparable analysis, DCF/LBO/three‑statement models, Excel audit, all 11 data connectors.

investment-banking – CIM, teaser, buyer list, M&A models, deal tracking for full‑cycle investment banking.

equity-research – Performance commentary, first‑cover reports, morning‑call minutes, investment‑logic tracking.

private-equity – Deal screening, due‑diligence checklists, IC memos, post‑investment monitoring.

wealth-management – Client review, financial planning, rebalancing, tax‑loss harvesting.

fund-admin – Reconciliation, accruals, rolling adjustments, NAV calculation for fund operations.

operations – KYC parsing, rule engine, compliance validation.

Partner plugins from LSEG and S&P Global add bond pricing, option volatility and funding‑digest capabilities.

Slash Commands for One‑Click Professional Output

/comps

– Comparable company analysis with transaction multiples. /dcf – DCF valuation with WACC and sensitivity analysis. /lbo – Leveraged‑buyout model for PE scenarios. /3‑statement‑model – Auto‑populate three‑statement model. /cim / /teaser – Generate confidential memoranda or teaser briefs. /earnings – Auto‑generate earnings commentary from filings and calls. /ic‑memo – Investment‑committee memo for PE/VC decisions. /tlh – Identify tax‑loss harvesting opportunities and avoid wash‑sale rules.

Data flow: upstream data → Claude processing → direct export to Excel, PowerPoint or Word, completing a closed‑loop workflow.

Zero‑Code Customization & Minimal Deployment

The repository is Apache‑2.0 licensed, written in plain Markdown and YAML, and requires no build steps. Teams can clone, modify, and deploy without specialized development skills.

Deployment Modes (Same Codebase)

Human‑in‑the‑loop (Plugin mode for Claude Code/Cowork)

# Add official plugin marketplace
claude plugin marketplace add anthropics/claude-for-financial-services
# Install core package (includes all 11 connectors)
claude plugin install financial-analysis@claude-for-financial-services
# Install desired agents (examples)
claude plugin install pitch-agent@claude-for-financial-services
claude plugin install gl-reconciler@claude-for-financial-services
claude plugin install market-researcher@claude-for-financial-services
# Install vertical plugins as needed
claude plugin install investment-banking@claude-for-financial-services
claude plugin install equity-research@claude-for-financial-services

Unattended (Managed Agent mode)

export ANTHROPIC_API_KEY=sk-ant-...
scripts/deploy-managed-agent.sh gl-reconciler

Both modes share the same system prompt and skill set.

Deep Office Suite Integration

Excel – Auto‑model from 10‑K/10‑Q, formula audit, sensitivity analysis.

PowerPoint – Charts auto‑update with data changes, accelerating deck creation.

Word – Generate credit memos and research reports following institutional templates.

Outlook – Inbox triage and tone‑aware auto‑replies, acting as a personal assistant.

Benchmark Performance

The underlying model is Claude Opus 4.7, which achieved 64.37 % on the Vals AI Finance Agent benchmark – the highest score reported for end‑to‑end financial‑scenario capability.

Domestic Deployment Recommendations

Replace data sources with local providers such as iFinD, Wind, Eastmoney, Juyuan, Hang Seng to cover A‑share and Hong‑Kong markets.

Swap the model for domestic large models (e.g., DeepSeek, Baichuan, Zhipu, iFlytek) while retaining the Agent framework to satisfy compliance.

Reuse the open‑source skill design, sub‑agent decomposition, command hierarchy, and audit‑log structure to reduce R&D cost.

Adapt for on‑prem deployment: keep data within the firewall, enable audit trails, and meet local regulatory standards.

Repository

https://github.com/anthropics/financial-services