§ 03.1 · Sentry

Sentry. Where research becomes a position.

Binomial's in-house research platform. The layer between raw data and a trade.

Routed specialists · Canvas · Tasks · Excel · SDK · BinomialHash compaction
§ 03.1.1 — The routing layer

One supervisor.
A desk of specialists.

Sentry's supervisor evaluates every query against a deterministic routing algorithm — matching topic, context and intent to the agent with the right tools. Each specialist is separately prompted and separately tooled — its own data sources, its own discovery patterns, its own answer style.

Every thread carries a workspace — charts, tables, models, code outputs — that agents read, write to and update as the conversation moves. The workspace is the shared state between human and machine; action results feed back into the next turn so the agent sees exactly what the analyst sees.

Memory is persistent and cross-surface. An analyst pins a peer group or a valuation preference once; it follows them into the next thread, the next Excel cell, the next scheduled Task. Context is managed, not repeated.

Equity research
Quant & risk
Derivatives
Capital flows
Market signals
Main · routingSupervisor
tag
Name
desc
Drag a specialist or supervisor to rearrange the network
listening  
§ 03.1.2 — The analysis layer

The analysis behind the trade.

The full research workflow — from risk decomposition to post-trade attribution. Every number carries its source, its timestamp, and the tool that produced it.

01 · Risk
Risk decomposition, not risk estimation.
Full-stack risk analytics built for portfolio-level decisions. Every metric computed from live positions against live data.
§ 03.1.3 — The surface layer

Same query.
Four surfaces.

The same agent fleet, the same routing, the same citation trail — accessed from the interface that fits the work.

AAPL deep-dive long_book +
S
Risk decomposition · long book vs SPY σ Risk
Factor exposures
Market
0.92
Size
0.34
Momentum
0.18
Value
−0.22
Quality
0.41
Positions · risk contribution
TickerWtβRisk σ
AAPL42%1.08+0.31
MSFT35%1.15−0.12
NVDA23%1.62+0.81
Cumulative return vs SPY · 90d
β SPY
1.12
VaR 95%
−3.2%
Track err
4.2%
Act share
68.1%
Sharpe
1.84
S
» Risk decomposition of the long book vs SPY
σ Risk
Pre-market brief Active daily · 06:00 ET · weekdays
Entrance
Start
End
Input
Portfolio
Files
Agent
σ Risk
§ Document
Comps
Output
PDF
Workbook
Deliver
Email
Slack
Start
Portfolio
Files
the long book
Prompt
» Risk decomposition vs SPY
σ Risk
PDF
#brief
next run: tomorrow 06:00 last: 14m ago · completed
fx =SENTRY.METRIC(A1, "evToEbitda")
ABCD
1 AAPL252.4117.3×−3.2%
2 MSFT448.2032.1×−1.8%
3 NVDA891.0561.8×−4.1%
Risk decomposition vs SPY
5 β SPY1.12
6 Track err4.2%
7 Act share68.1%
Sentry
S
» Risk decomposition of the long book vs SPY
Decomposing 3 positions against SPY. Writing β, tracking error and active share to rows 5–7.
Analyzing…
python3
>>> from sentry import Client
>>> c = Client(api_key="…")
>>> c.query("Risk decomposition of the long book vs SPY")
→ streamed response, structured output
S.01 · Canvas
The workspace.
Tabbed analysis environment. Datasets, charts, pivots, tool workflows — and agents that modify the workspace directly: pull a peer group, build the comp table, decompose the risk. Every tab persists; every artifact exports to PDF, PowerPoint or Excel.
cross-asset joinsstatistical testingcomputational lineagemulti-format export
Sentry is built in-house, deployed in-house, and accountable to the book. It is the way Binomial converts research into capital — and the surface where humans and machines meet.