If you still think AI Agents are limited to “help me draft an email,” this Anthropic repo will recalibrate your expectations.
Last week, Anthropic quietly launched financial-services on GitHub—and within a week it garnered 23,400 stars, with 9,480 added just this week. This is not a toy project: it contains nine fully functional Agent templates designed for real-world financial workflows—from investment banking pitch books to LP reconciliation audits for funds.
What’s Inside
The core idea behind this repository is refreshingly straightforward: break down the most repetitive, time-intensive tasks in finance into named, purpose-built Agents—each equipped with its own skills and data connectors, ready to deploy out-of-the-box.
| Agent | Purpose |
|---|---|
| Pitch Agent | Generates branded pitch decks end-to-end from comps, precedents, and LBO models |
| Meeting Prep Agent | Produces pre-meeting briefing packs for clients |
| Market Researcher | Transforms industry/theme inputs into industry overviews, competitive landscapes, and peer comparisons |
| Earnings Reviewer | Analyzes earnings calls + filings → updates models → drafts research reports |
| Model Builder | Runs DCF, LBO, three-statement, and comps models directly in Excel |
| Valuation Reviewer | Ingests GP packages, executes valuation templates, and prepares LP reports |
| GL Reconciler | Identifies discrepancies, traces root causes, and routes approvals |
| Month-End Closer | Handles accruals, carryforwards, and variance explanations |
| KYC Screener | Parses account-opening documents, runs rule engines, and flags gaps |
Each Agent is a self-contained plugin—installing one means installing all required skills automatically, with no additional configuration needed.
Two Deployment Options—One Codebase
The most practical design feature of this repo is “write once, run anywhere”:
Option 1: Claude Cowork Plugin. In Cowork, go to Settings → Plugins → Add Plugin, paste the repo URL, and select the Agents you need—or upload a zipped subdirectory directly. Claude Code users can do it in one command:
claude plugin marketplace add anthropics/financial-services
claude plugin install pitch-agent@claude-for-financial-services
Option 2: Claude Managed Agents API. Ideal for teams with their own workflow orchestration engines. The repo includes complete agent.yaml definitions, leaf-worker sub-agent configurations, and steering-event examples. A single deployment script auto-resolves file references, uploads skills, creates sub-agents, and POSTs to /v1/agents:
export ANTHROPIC_API_KEY=sk-ant-...
scripts/deploy-managed-agent.sh gl-reconciler
Same system prompts, same skills—you choose where they run.
Partner Ecosystem Already Onboard
The repo includes a plugins/partner-built/ directory, already featuring plugins from LSEG and S&P Global—meaning third-party data providers are embedding their data connectors directly into Claude’s Agent ecosystem. This isn’t Anthropic going solo; it’s the entire industry contributing.
Let’s Be Honest
Anthropic includes a disclaimer in the README: "Nothing in this repository constitutes investment, legal, tax, or accounting advice." These Agents produce drafts—every step still requires human review and sign-off. This is not an autonomous trading bot; it’s an assistant built to eliminate rote labor for analysts.
That said, when a Pitch Agent can generate a branded pitch deck from comps, precedents, and LBO models with one click—saving junior analysts 20 hours per week—that’s transformative in investment banking. Industry insiders understand the implications better than I ever could.
Is This Right for You?
- If you work in financial services technology, spend an afternoon reading through this repo to identify which Agents can be embedded directly into your existing workflows.
- If you’re an Agent developer, the orchestration patterns in
managed-agent-cookbooks/(orchestrator → leaf-worker sub-agents → handoff event routing) provide an excellent reference architecture. - If you’re outside finance, this repo itself serves as a best-practice blueprint for “industry-specific Agent templates”—how to name them, how to package skills, and how to support both plugin and API deployment simultaneously.
The repo also includes scripts/orchestrate.py, demonstrating how to route handoff_request events within your own orchestration layer. This design deserves dedicated study.