A live knowledge graph built from the information you already have

Investor letters, transcripts, emails, factsheets — Terrace continuously structures your information flow into a unified, queryable graph. Reconciled entities, resolved relationships, clean time series. Ready for analysis, APIs, or agents.

Request access

The problem

Allocators accumulate thousands of investor letters, call transcripts, DDQs, and emails every year. The information is already there — returns, exposures, team changes, positions — but it is locked inside unstructured formats, scattered across inboxes and shared drives. No database you can buy contains what your own information contains.

How it works

Ingest

PDFs, transcripts, email bodies, and data files flow into Terrace — uploaded directly or integrated into your existing workflow.

→

Structure

LLM extraction, entity resolution, deduplication, and audit. Messy, inconsistent information becomes a reconciled knowledge graph with unified entities and clean time series.

→

Query

Browse the graph directly, hit the API, or connect your own agents via MCP. The data is structured for machines, not just humans.

What the graph gives you

Unified return series

Monthly returns across every manager, reconciled from multiple sources. Share class tracking, observation audit trails, conflict resolution.

Exposure history

Sector, geography, and asset class exposures over time. Cross-category validation. Normalised to GICS and standard taxonomies.

Entity resolution

The same fund appears under 5 names across 20 documents. Terrace resolves them into one entity and remembers for next time.

Team and career graph

Portfolio managers, analysts, and their career histories — extracted from letters, brochures, and filings. Linked across firms.

Semantic search

Vector and full-text search across every piece of information ever ingested. Find what was said about a sector, a position, a risk event.

Continuous quality audit

Deterministic checks and LLM-powered deep audit catch duplicates, outliers, misclassifications, and data conflicts automatically.

The graph compounds

Firm Highgate Meridian Capital ├── Strategy Asia-Pacific Long Short Equity │ ├── Vehicle HMC APAC Master Fund LP │ │ ├── returns: 48 months, 3 share classes │ │ ├── exposures: sector, geography, market cap │ │ └── positions: 47 holdings, FIGI-resolved │ └── Vehicle HMC APAC Offshore Feeder Ltd ├── Strategy Japan Focus │ └── Vehicle HMC Japan Master Fund LP ├── Person D. Nakamura, PM — prev. multi-manager, macro fund ├── Person L. Wei, Analyst — prev. long-only, event-driven └── 12 documents, 3 IAPD brochures, 1 13F filing resolved from: “Highgate Meridian Capital”, “Highgate Meridian Capital Pte. Ltd.”, “HMC”, “Highgate Meridian Cap.” across 12 documents over 4 years

Access the graph

Browser

Browse entities, return grids, exposure charts, team pages, and transcripts. Search across everything.

API

Structured endpoints for every entity type, time series, and search. Build your own dashboards and reports.

Agents

Connect via MCP or CLI. Your agents query pre-structured data instead of parsing raw sources on the fly.