Assistant

"Your edge shouldn't depend on whether someone built an API for it."

The Problem Assistant Solves

Every agent on your desk runs on data. Vanta reads regime signals. Luma aggregates cross-venue intelligence. Meridian scans macro feeds. But all of them depend on structured, well-maintained data sources with clean APIs, exchange feeds, on-chain indexers, news aggregators, economic calendars.

The reality is that some of the most valuable trading data in the world doesn't come from a clean API. It lives on a government website that updates a PDF every Thursday. It's buried in a Telegram channel where a whale tracker posts screenshots. It's a niche forum where options flow data gets shared in raw text. It's a regional news outlet that covers mining regulations before anyone else picks it up. It's a spreadsheet that a DeFi protocol publishes on their docs page with no programmatic access.

Traders see this data. They know it has value. But integrating it into their workflow means manually checking dozens of sources, copying numbers into spreadsheets, and hoping they don't miss an update at 3am.

Assistant exists to close that gap. Its entire job is to go out, find the data you care about regardless of where it lives or how messy it is, scrape it, normalize it, and deliver it to your desk in a format every other agent can use.

Why This Matters

The best traders have always had an information edge. Not insider information, just better data, gathered faster, from more places, with more context. In Market Wizards, Jack Schwager found that the traders who consistently outperformed didn't have secret formulas. They had better inputs. They read more, tracked more, and connected dots that others missed because they were looking at sources nobody else bothered to check.

The problem is that maintaining that edge manually doesn't scale. You can track five niche sources by hand. You can't track fifty. And the moment you stop checking one of them, that's usually when the signal fires.

Assistant scales your information edge. You tell it what to watch, and it watches. Continuously. Reliably. And it translates what it finds into structured intelligence that Luma can score, Riven can factor into strategy, and Meridian can weigh against the macro picture.

Real-World Scenarios

Scenario 1: Tracking Unreleased Mining Data

A trader notices that a provincial government in China publishes monthly energy allocation reports as PDFs on their website. These reports contain hydroelectric output data that correlates with Bitcoin mining hashrate changes 2-3 weeks before on-chain metrics reflect them.

No API exists for this. No data provider covers it. But the trader knows it's valuable.

They point Assistant at the URL and say: "When a new PDF appears, extract the hydroelectric output table and calculate month-over-month change. If output drops more than 10%, flag it as a potential hashrate disruption."

Assistant checks the page daily. When a new report drops, it extracts the data, normalizes it, and pushes a structured alert to the desk. Meridian factors the hashrate risk into its systemic assessment. Riven adjusts strategy to account for potential mining-related sell pressure. The trader has a 2-week lead on everyone using standard on-chain data.

Integration With the Desk

Assistant is the only agent whose job is defined entirely by you. Every other agent has a fixed role: Vanta reads regimes, Luma scores signals, Brix manages risk. Assistant's role is whatever data gap you need filled.

This makes it the most personal agent on your desk. Two traders using PerpDesk might have completely different Assistant configurations because they trade differently, follow different sources, and see edge in different places. One trader's Assistant might be tracking government energy reports and whale wallets. Another's might be monitoring options flow screenshots from a Discord server and TVL changes across five DeFi protocols.

Your edge is your own. Assistant makes sure your desk has access to it.