“Just point an AI at our email.”
A pile isn't intelligence.
The inbox, resolved: companies, metrics, and touchpoints, cited and permissioned, over MCP and a REST API. Memory worth pointing anything at.
The pile breaks every project.
Wire anything to raw email and you hit the same seams.
One curated layer. Every surface you build.
Four ways to build without owning ingestion or retrieval.
The firm's AI tools, portfolio-aware in minutes.
Connect Claude, ChatGPT, or your own agents over MCP. Every task gains structured records, cited answers, and the app's permissions.
Cited time-series, straight into the warehouse.
Companies, funds, notes, meetings, connections, and full tracker value histories, over a versioned REST API with an OpenAPI spec. Webhooks fire the moment new communication is indexed, so nothing has to poll.
Browse the API reference ↗Cited internal tools, without building retrieval.
Hybrid search, block-level semantic and full-text scored together, across everything indexed, decks behind links included. Retrieval, ranking, and provenance come with the layer.
Provenance and permissions travel with the data.
Every object carries source attribution and access rules. Partners' sharing levels and member visibility govern what any surface returns, so your tools inherit the firm's governance.
Provenance is a field, not an afterthought.
In the app, citations are chips you click. In the API, a field: every extracted value carries a citations array naming the element, file, or email it came from, each source naming its parent back to the email it arrived on.
Questions builders ask us.
Build on memory, not on a pile.
Connect an inbox, generate a key, and point your first tool at cited portfolio data. Curation, provenance, and permissions, already done.