The AI workspace for investment research

Turn market noise
into signal clarity.

50+

Workflow nodes

34

Widget types

6

Frontier AI labs

100%

Runs audited

01 Flagship · The Conviction Engine

Five specialists.
One committee. Every ticker.

Five AI specialists analyze every name on your watchlist. They debate. The committee synthesizes a single verdict — and tells you which specialist disagrees, and why. It remembers yesterday's call and only alerts you when conviction shifts.

orbital — conviction engine
Live

Visible-workflow architecture

Every fetcher, every specialist, every vote — on the canvas, editable. No sealed black box.

Dissent, surfaced

The committee tells you which specialist disagrees and why, so you see the bear case before you size the position.

Memory between runs

It remembers yesterday's verdict and alerts you only when conviction flips or shifts — not every morning.

02The research loop

Build it once.
Run it on everything.

Drag data, transformation, AI, and alert nodes onto a canvas. Wire them together. Point the workflow at a watchlist and every symbol gets the same disciplined treatment — on a schedule, with every run audited.

WATCHLISTMomentum-1212 symbols · defaultFUNDAMENTALSEarnings · guidance · ratiosMARKET DATALive prices · technicalsINSIDER + GOV TRADESForm 4 · congressionalAGENTRisk AssessorJSON output · 4 toolstemp 0.2OUTPUTDashboard widgetALERTSlack · on change
Workflow · daily-watchlist-run Run #1,284 · branches in parallel

01

Watchlist

Unlimited symbols with live snapshots, sections, flags, and rich notes. One list feeds every workflow.

02

Workflow

A visual pipeline of ~50 node types — data, transforms, AI, metrics, alerts. Branches run in parallel.

03

Agents

Reusable AI analysts with instructions, tools, and structured output. Configure once, call from any workflow.

04

Dashboard

Bind outputs to 34 widget types on a live grid. Values update run after run, with provenance one click away.

03K-LLM synthesis

Never trust
a single model.

Every model has blind spots. Orbital runs the same question through K frontier models in parallel, then synthesizes the best answer — with the dissenting view surfaced as a first-class output, not hidden behind one model's confidence.

ClaudeGPTGeminiGrokDeepSeekPerplexitySYNTHESIZERK-LLM consensussame prompt · K modelsdissent surfacedCONSENSUSBUY · 78% agreementDISSENT1 of 6 models flags ratesensitivity in the thesis —audit per-model reasoning →

Bring your own keys — Claude · GPT · Gemini · Grok · DeepSeek · Perplexity

04Provenance

Every number has
a paper trail.

No black boxes. Click any value on any dashboard and trace it back to the workflow run, the node, and the exact AI execution that produced it.

Widget value

PLTR · 82

Conviction scoreboard

Workflow run

#1,284

Today · 06:00 ET · 47s

Node result

synthesizer

Committee verdict

Agent execution

Risk Assessor

Model + prompt + tokens

05The full platform

Institutional tooling,
without the institution.

Alternative data, included

Insider trades, congressional trading, analyst dispersion, earnings surprises — terminal-grade data as drag-and-drop nodes.

Alerts that respect you

Slack, Discord, email, SMS, or webhook — with content-aware dedup and cooldowns so you only hear about real change.

Scheduled autonomy

Publish a workflow and put it on cron. Your research runs before the open, every day, whether you're watching or not.

Semantic memory

Workflows write to and search a persistent memory, so each analysis builds on prior conclusions instead of starting cold.

A real escape hatch

No-code for everyone — plus a sandboxed code node with statistical and financial helpers when you need full control.

Your models, your keys

Bring your own keys for six frontier labs, or use Orbital credits. Swap models per agent, per node, per question.

Start turning noise
into signal.

Orbital is in private beta. Join the waitlist for early access to institutional-grade research intelligence.