MCP servers
MCP server for fund portfolios (cross-border holdings)
Two funds in two countries can hold the same company and describe it in ways that share no common field. One reports a CUSIP quarterly to the SEC; the other reports a Portuguese asset-class label monthly to CVM. Comparing them is less an analysis problem than an identity problem.
This server presents both under one portfolio schema. A position is a position: an issuer, a quantity or weight, a period, and the date the disclosure became public. Where the underlying regimes genuinely differ — reporting thresholds, asset-class taxonomies, disclosure lags — the differences are exposed as fields rather than smoothed away, because a comparison that hides them is worse than no comparison.
For an agent this means a cross-border exposure question is a single tool call with two jurisdictions in scope, and the answer arrives with the disclosure dates that determine whether it was knowable at all.
hindcast — session
previewhindcast> which funds disclosed exposure to this issuer, either market?
→ tool: fund_portfolios { entity:"...", markets:["us","br"], as_of:"..." }
✓ disclosed positions with fund, weight and disclosure date
source: SEC 13F and CVM CDA filings, one schema
What's covered
- Tool: portfolio lookup by fund across US and Brazilian regimes
- Tool: reverse lookup — which funds disclosed a given issuer
- One position schema with regime differences exposed as fields
- Disclosure date on every position, separate from the period
- Issuer identity resolved through the shared entity graph
- Reporting thresholds and taxonomies labeled, never silently merged
Join the waitlist — this page's dataset is already selected.
Questions
Are US and Brazilian disclosures really comparable?
Partly, and the tool is explicit about where they are not. Thresholds, frequencies and asset-class definitions differ; keeping those visible is what makes the comparable parts trustworthy.
Can I see a fund's history rather than one snapshot?
Yes — portfolios are stored as dated observations, so the sequence of disclosures is queryable and changes between them are derivable without reconstructing anything by hand.
Related
Join the waitlist. Vote the roadmap. First in, first served.