Large financial teams face two parallel problems: (1) turning model outputs into auditable, repeatable analyst work product, and (2) wiring LLMs to licensed market data and internal systems safely. This repository addresses both by bundling end‑to‑end agent templates, vertical skill bundles, and MCP connectors so firms can adopt Claude-driven workflows while keeping human review and compliance at the center.
What Sets It Apart
- Opinionated, workflow-first templates: named agents (Pitch Agent, Market Researcher, GL Reconciler, etc.) map to real FSI tasks so teams get a usable starting point rather than a blank prompt. This reduces time-to-prototype and helps maintain consistent outputs across contributors.
- Shared skill library + connector fabric: vertical plugins centralize modeling skills (comps, DCF, LBO, 3-statement) and MCP integrations, so multiple agents reuse the same logic and data endpoints—helpful for governance and testing.
- Dual deployment surface: each agent ships both as a Cowork plugin (interactive, in-product) and a Managed Agent template (headless/orchestrated). That lets teams choose interactive analyst workflows or automated pipelines without reauthoring skills.
- Operational focus, not investment decisioning: templates explicitly stage outputs for human sign-off and include security/steering notes—suitable for firms that need audit trails and control over data flows.
Who It's For & Trade-offs
Great fit if you operate in a regulated finance environment and need repeatable, auditable LLM workflows that integrate licensed data providers. Teams with existing MCP subscriptions or enterprise data vendors will find the connector list and skill templates especially valuable. It’s also useful for product teams building internal analyst tooling around Claude.
Look elsewhere if you need a ready-made, fully managed trading or execution system (the repo does not execute trades or provide compliance approvals), if you cannot obtain required MCP/data vendor access, or if you require an LLM-agnostic stack—while the architecture is file-based, it is opinionated around Claude/Anthropic tooling and MCP connectors.
Where It Fits
Consider this repository a middle layer between enterprise data providers and analyst UI/orchestrator: the repo supplies the agent logic, reusable commands (/comps, /dcf, /earnings, etc.), and connector manifests; your firm supplies credentials, governance, and the final approval gates. That makes it a pragmatic integration template rather than a drop-in product.
