Gradient CIO

Institutional portfolio analytics in Claude: capital market assumptions, portfolio simulation, manager due diligence, DDQ review, and daily macro research.

Hosted MCP Server

npx add-mcp 'https://mcp.gradientcio.com/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

  • FRED — Federal Reserve Bank of St. Louis
  • Federal Reserve — Board of Governors of the Federal Reserve System
  • U.S. Treasury — U.S. Department of the Treasury
  • SEC EDGAR — Form ADV · 13F · Form D
  • CFTC — Commodity Futures Trading Commission
  • IMF — International Monetary Fund
  • World Bank
  • Shiller (Yale) — Robert Shiller CAPE and long-run market data at Yale University

Illustrative example

Run simulations, and AI can help evaluate the results.

What range of outcomes should we plan for?

Simulation paths

1,000

Selected horizon

10 years

The median path compounds to 1.76x, but compounding stretches the upside far more than the downside, so the average outcome flatters what a typical path delivers. The binding constraint is the left tail: a 95% one-year CVaR of −16.4% against a 5.8% median return, a Sharpe ratio of only 0.24. Set spending and liquidity buffers against the lower quartile, and test whether trimming equity beta narrows that tail more than it costs in return.

If commodity prices rise 20% over the next three years, how would that change the shape of this simulation, given our portfolio commodity exposure?

Median after commodity rise

1.79x

Wider range at year 3

+4 pts

Less at the center than in the tails. Real assets (12.6% of NAV) and energy held through funds (3.2%) gain, while margin pressure on public equity and higher yields on fixed income offset most of that gain. The median ten-year outcome barely moves, from 1.76x to about 1.79x. The range widens instead: by year three the 5th-to-95th-percentile band is about 4 points wider, and slightly more paths finish below the start. Real assets sit 2.4 points under the 15% policy target, so the portfolio is less hedged than policy. Closing that gap would turn a commodity rally into a modest tailwind.

Which exposures would hurt most in a stagflation shock, and does anything we hold offset it?

Return in a stagflation year

−11.6%

Median after the shock

1.56x

Stagflation costs the portfolio 11.6%, against 10.2% for policy. The gap comes from exposures, not volatility: the value tilt carries 38% of factor risk and depends on cyclical earnings that inflation squeezes. The short-duration position works the other way, which is why the portfolio beat policy in the 2022 rates shock. A stagflation year would lower the median ten-year outcome to about 1.56x. Test whether trimming the value tilt narrows that gap more than it costs in return.

The platform

Four capabilities across the investment process.

Holdings, market research, risk models, and committee papers live in one workspace, so every step of an allocation decision builds on the last.

Portfolio Intelligence

One normalized view of the whole portfolio: allocations, commitments, holdings, liquidity, and private-market cash flows.

Market and Manager Research

Macro research, capital market assumptions, and manager diligence drawn from dated primary sources.

Quantitative Analytics

Factor risk, stress tests, simulation, and attribution calculated on documented models and assumptions.

Investment Decision Support

Allocation studies and committee papers that show the calculation, source, and assumption behind every figure.

Illustrative example

Test an allocation against its policy benchmark before you commit.

What are the odds the candidate allocation beats the policy benchmark over three years versus five?

Odds of beating policy over 5 years

74%

Narrower range at 5 years than 3

21%

Good odds at both horizons, and better at five: 70% of paths beat policy over three years and 74% over five. The median edge holds at close to a point a year either way; time changes the spread, not the center. With 3.0% tracking error, the 10th-to-90th-percentile range of annualized excess return is 21% narrower at five years, and the bottom decile improves from −1.3 to −0.9 points a year. A three-year shortfall of a point or so is ordinary noise for this allocation. Set the review horizon and the tracking-error tolerance against the five-year band, so the committee does not reverse the decision on a result the plan already expects.