Anthropic Releases Finance Agent Templates for Claude Cowork, Claude Code, and Managed Agents

Anthropic has released ten ready-to-run finance agent templates for Claude Cowork, Claude Code, and Claude Managed Agents. The templates target work such as pitchbooks, KYC screening, market research, model building, valuation review, general ledger reconciliation, and month-end close. The useful point for business teams is that Anthropic is packaging repeatable analyst workflows with plugins, cookbooks, connectors, and Microsoft 365 add-ins rather than selling a generic chatbot for finance.

Key takeaways

• Anthropic says the release includes ten finance agent templates that ship as Claude Cowork and Claude Code plugins and as cookbooks for Claude Managed Agents.
• The public anthropics/financial-services repository lists the agent plugins, vertical plugins, managed-agent cookbooks, partner-built plugins, connector configuration, and Microsoft 365 provisioning tooling.
• Claude add-ins for Excel, PowerPoint, Word, and Outlook are positioned as a way to carry context between models, spreadsheets, decks, documents, and email.
• Anthropic names partner connectors and MCP integrations for finance data and workflow providers including Dun & Bradstreet, Fiscal AI, Financial Modeling Prep, Guidepoint, IBISWorld, SS&C Intralinks, Third Bridge, Verisk, and Moody's.
• VALS lists Claude Opus 4.7 as the top model on its Finance Agent benchmark at 64.37% accuracy, which supports Anthropic's benchmark claim but does not remove the need for human review.

Why it matters

Finance teams often have well-defined workflows but messy inputs: filings, transcripts, data rooms, Excel models, emails, policies, and approval trails. Anthropic's release is notable because it maps agents to named jobs and review boundaries, not just to broad "AI assistant" use cases. A firm could start with a narrow workflow such as KYC file assembly or month-end variance commentary, then decide whether it belongs as an analyst-side plugin or as a managed agent behind an internal workflow engine.

A practical rollout should begin with one low-regret workflow where source documents are available, the review owner is obvious, and the final action is staged rather than automatic. Good candidates are meeting briefs, first-pass earnings summaries, pitchbook drafting, or reconciliation exception packages. Avoid starting with workflows that approve onboarding, execute trades, post ledgers, or send client-facing material without sign-off.

What to verify before you act

Check whether your firm can use the required Claude products, plugins, managed-agent APIs, Microsoft 365 add-ins, and data connectors under existing vendor and compliance rules. Review the GitHub repository's per-agent materials before installation, because the templates are starting points that need tuning for modeling conventions, risk policy, approval flow, and source access. Treat the VALS benchmark as a capability signal only; you still need task-level evaluation with your own filings, templates, data vendors, spreadsheets, and professional reviewers.

Source check

Anthropic's announcement confirms the ten templates, Claude Cowork and Claude Code plugin path, Managed Agent cookbook path, Microsoft 365 add-ins, partner connector list, and stated Finance Agent benchmark claim. The official Anthropic GitHub repository corroborates the concrete package structure, named agent plugins, vertical plugins, managed-agent cookbooks, connector descriptions, and review caveats. VALS independently lists Claude Opus 4.7 at 64.37% on the Finance Agent benchmark, corroborating the benchmark number used as context rather than as a deployment guarantee.

For broader workflow selection, LinkLoot's AI agent tools guide can help teams separate useful agent workflows from risky automation shortcuts.