Claude Enters Finance: 10 Open‑Source Financial Agent Templates Unveiled

Release Overview

Anthropic released ten open‑source financial Agent templates that cover investment‑banking, credit, compliance and accounting workflows. Each template bundles three components:

Skills : task instructions and domain knowledge.

Connectors : governed access to data sources such as S&P Capital IQ, PitchBook and Moody’s.

Subagents : Claude models that execute subtasks (e.g., comparable‑company selection, methodology validation).

Deployment Options

Agents can be used in two modes:

As a Plugin installed in Claude Cowork or Claude Code for analyst‑desktop collaboration.

As a Cookbook running on Claude Managed Agents for unattended batch processing.

All assets are released under the Apache 2.0 license in the GitHub repository anthropics/financial-services.

Repository Structure

plugins/agent-plugins

– ten self‑contained agent templates. plugins/vertical-plugins – vertical‑specific Skill + Command + Connector bundles. plugins/partner-built – partner‑authored plugins (LSEG, S&P Global). managed-agent-cookbooks – deployment templates for Managed Agents. claude-for-msft-365-install – IT‑admin tooling for the Microsoft 365 add‑in. scripts/ – deployment, validation and orchestration scripts.

Agent Matrix

Agents are grouped into three vertical domains:

Research & Client Coverage (front‑office) : Pitch builder, Meeting preparer, Earnings reviewer, Model builder.

Credit, Risk & Compliance (middle‑office) : Market researcher, KYC screener.

Finance & Operations (back‑office) : Valuation reviewer, General ledger reconciler, Month‑end closer, Statement auditor.

Slash‑Command Matrix (excerpt)

/comps

– comparable‑company analysis + transaction multiples. /dcf – DCF valuation with WACC and sensitivity analysis. /lbo – leveraged‑acquisition model. /3‑statement‑model – one‑click three‑statement model fill. /debug‑model – Excel model audit (formula tracing, hard‑code detection, balance check). /cim – draft Confidential Information Memorandum. /teaser – anonymous one‑page company teaser. /buyer‑list – strategic and financial buyer universe. /merger‑model – deal‑level accretion/dilution analysis. /earnings – earnings commentary. /initiate – institution‑level initial coverage report. /morning‑note – morning‑note summary + trade ideas. /ic‑memo – investment‑committee memorandum. /dd‑checklist – due‑diligence checklist grouped by workflow. /unit‑economics – ARR cohort, LTV/CAC, net retention, revenue quality. /returns – IRR/MOIC sensitivity table. /client‑review – client‑meeting preparation. /tlh – tax‑loss‑harvesting detection + wash‑sale protection.

Installation via Claude Code

# 1. Add marketplace
claude plugin marketplace add anthropics/claude-for-financial-services

# 2. Install core package (required)
claude plugin install financial-analysis@claude-for-financial-services

# 3. Install specific agents as needed
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

# 4. Install vertical plugins
claude plugin install investment-banking@claude-for-financial-services
claude plugin install equity-research@claude-for-financial-services

For Managed‑Agent deployment, set the Anthropic API key and run the provided script, e.g.:

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

Two Deployment Modes, Same Code

Plugin in Claude Cowork or Claude Code – human‑AI collaboration on the analyst’s desktop.

Managed Agent – runs on Anthropic’s platform to process whole deal‑books overnight without supervision.

Writing the template once enables both modes, halving engineering effort.

Office Suite Integration

Claude is packaged as an official Microsoft 365 add‑in for Excel, PowerPoint, Word and Outlook. Context automatically flows between the apps: a DCF model built in Excel appears in a PowerPoint deck, and Outlook can draft cover notes in the analyst’s voice.

Performance Benchmark

The agents run on Claude Opus 4.7, which Anthropic claims is state‑of‑the‑art for financial tasks. On the Vals AI Finance Agent benchmark the model achieved 64.37 % , the highest score reported for end‑to‑end agent capability in finance scenarios.

Connector Endpoints (MCP)

The core financial-analysis plugin lists eleven publicly documented MCP endpoints. Selected examples:

Daloopa – https://mcp.daloopa.com/server/mcp Morningstar – https://mcp.morningstar.com/mcp S&P Global – https://kfinance.kensho.com/integrations/mcp FactSet – https://mcp.factset.com/mcp Moody’s – https://api.moodys.com/genai-ready-data/m1/mcp MT Newswires – https://vast-mcp.blueskyapi.com/mtnewswires Aiera – https://mcp-pub.aiera.com LSEG – https://api.analytics.lseg.com/lfa/mcp PitchBook – https://premium.mcp.pitchbook.com/mcp Chronograph – https://ai.chronograph.pe/mcp Egnyte – https://mcp-server.egnyte.com/mcp Access requires a subscription or API key, but the endpoints are publicly documented for direct Agent integration.

Pros and Cons

Pros

Vertical specialization provides a complete SOP for investment‑banking and asset‑management tasks.

Deep integration with the Office suite where analysts spend ~90 % of their time.

End‑to‑end audit‑ready workflow (templates + connectors + Managed Agents + Skills).

Customer endorsements from firms such as Citadel, Carlyle and BNY.

Cons

Underlying Opus 4.7 model remains closed source.

Compliance requirements for Chinese financial institutions may limit direct Claude usage.

Connectors are primarily Western data sources; coverage for A‑share and Hong Kong markets is minimal.

Enterprise‑scale rollout still depends on further engineering for large‑scale deployment.

Conclusion

Anthropic’s strategy focuses on high‑value verticals rather than competing with generic models. Finance is the first battlefield because of high ticket size, standardized workflows and alignment with Claude’s safety‑first stance. The anthropics/financial-services repository serves as a concrete product‑planning blueprint for building AI agents in financial institutions.