Improvement
"Which positions are underwater, and why?" used to mean pulling holdings, rounds and marks from three places and lining them up by hand. The PortfolioIQ MCP now gives Claude every data point in PortfolioIQ, so the answer comes back as a live dashboard.
What Claude can now read
Metrics with their source document, as-of date and restatements. Holdings by fund and security, transactions, valuation history, funding rounds, and full cap tables with every shareholder. Fund performance and terms, documents, AI insights from board decks, notes, and data collection requests with founders' responses.
The connection is read-only. Claude can analyse your data; it cannot change it.
What you can now build
We took real questions a GP actually ask and answered each as a live dashboard in Claude, using only the MCP. All figures below come from a demo firm with 20 portfolio companies and two funds, not a customer.
Underwater positions: Every position marked below cost, at security level, with the record beside it: entry price, later round prices, valuation marks, realised proceeds. In the demo, 4 of 59 positions were below cost. One traced to a later round priced under the firm's entry, $14.20 a share then $12.00.
Ownership: Your stake by fund, every other meaningful holder, the option pool, and which co-investors recur across the portfolio. Cap tables older than the latest round get flagged.


Exit scenarios: Recompute fund returns if your largest positions exit at 1x, 2x or 3x and a weak one is written off. The demo fund goes from 1.31x gross today to 2.01x at 2x and 2.71x at 3x.
Board-deck claims: Check every positive claim PortfolioIQ extracted from board decks against the metric it is about. Of 211 claims in the demo, 61 were supported and 22 contradicted.
Burn multiple: Computed live across the SaaS book from net burn and net new ARR, on each company's own fiscal quarters. Where ARR shrank, it shows "not meaningful" rather than a misleading number.
Notes and data collection: Which companies have no risk note (5 of 20 in the demo). What is still missing to close the quarter: 180 overdue requests across 12 companies, ranked by who to chase first.
Why the data layer is the point
Claude never had to pick the right source, line up fiscal years, map metric names or convert currencies. PortfolioIQ had already done it.

Verified: Every value ties to a source document, down to the page or spreadsheet cell, with an as-of date. Restatements keep the old value. When two documents disagree, you see both: a board deck said $3,020,000 of quarterly revenue, a spreadsheet said $3,024,000.
Normalized: Three demo companies report in euros or rupees. Read raw, one ranks third of ten on ARR. In the firm's currency it is last, at $0.55M. Without normalization, the leaderboard is simply wrong.
Structured: Each company's own metric names map to one shared set of definitions, so the same question works across all 20 companies.
Get started
Connect PortfolioIQ to Claude at app.portfolioiq.co/mcp.



