Case study — Bespoke

Bespoke

An AI copilot for wealth advisors, built on the Claude Agent SDK. Advisors chat naturally to research SEC filings, query client portfolios, run quant analysis (RSI, MACD, risk metrics), and execute paper trades — powered by domain-specific MCP tool servers and specialized subagents (research, data analyst, quant, audit) over an AWS S3/Athena data lake. Battle-tested in production, then extracted into a reusable agent framework with memory, skills, scheduled agents, and deck generation out of the box.

Client

Bespoke

Sector

AI Agents · Claude Agent SDK · FastAPI · Next.js · AWS Athena · Python

Bespoke: AI Copilot for Wealth Management

Bespoke: AI Copilot for Wealth Management

The production AI copilot that turns natural conversation into research, analysis, and trades.

Introduction

Bespoke is an AI copilot built for wealth advisors on the Claude Agent SDK. Instead of jumping between terminals, spreadsheets, and filing databases, advisors simply chat — and the copilot researches SEC filings, queries client portfolios, runs quantitative analysis, and even executes paper trades on their behalf. It was battle-tested in production, then extracted into a reusable agent framework that ships memory, skills, scheduled agents, and deck generation out of the box.

Demo

What it does

  • Conversational research — pulls SEC filings, fundamentals, and news into a single answer.
  • Portfolio intelligence — queries client holdings, positions, and summaries on demand.
  • Quant on tap — computes technical indicators (RSI, MACD) and risk metrics without leaving the chat.
  • Paper trading — simulates trade execution so advisors can test ideas safely.
  • Compliance-aware — a dedicated audit agent verifies data and reviews for compliance.

Under the hood

The copilot runs on a modular architecture: domain-specific MCP tool servers (data lake, client data, portfolio, paper trading) feed specialized subagents — research, data analyst, quant, and audit — that coordinate over an AWS S3 / Athena data lake. Every capability is a swappable Feature, so the same framework powers a hello-world agent or a full production copilot.

Tech Stack

AI Agents · Claude Agent SDK · FastAPI · Next.js · AWS S3 / Athena · Python

Where is the same work happening in your business?

The assessment finds it, prices the fix, and tells you what it returns. Two to four weeks, fixed fee, quoted on the call. You keep everything it produces.