Vektor
Now in beta · Alpaca-powered

The agent that
trades with you.

It shows its work. It asks first.

Vektor runs every signal through a five-node reasoning pipeline and shows you the whole chain. Today, every trade waits for your approval. Bounded automation is next: the agent will act inside limits you set and flag anything outside them.

Early access coming soonSee how it works
ScoutScans the universe and picks candidates on autonomous runsSupervisorChecks market status and your portfolioResearchReads news and SEC filings, scores sentimentStrategyForms the signal and confidenceRiskApplies your limits in deterministic code
vektor · agent run · autonomous scan
[node 0]Scout · scanned 312 tickers · selected AAPL
[node 1]Supervisor · market open · portfolio fetched
[node 2]Research · 12 news items · 1 SEC filing (8-K) · bullish (0.73)
[node 3]Strategy · ensemble: BUY 0.81 · news: bullish 0.73
[node 4]Risk · position sizing: $1,240 · gate: APPROVED
[result]Queued for approval · confidence 0.81 · est. value $1,240
AAPLBuy
81%
Confidence
Buy 6 shares @ ~$206.67≈ $1,240
5
Pipeline nodes
596
Automated tests
0
Trades without a reasoning chain

Our mission

Built for the ones who weren't handed the playbook

Vektor brings agentic trading to mainstream retail investors: the next generation who were told this world requires an insider's resume. We think it requires something else entirely. Clear reasoning, honest guardrails, and a fair shot at learning.

Opening the door is not enough on its own. Easy access without guardrails is just a casino with better branding. So Vektor pairs the two: approachable enough for a complete beginner, and never reckless with anyone's money.

Literacy first

Financial literacy is the product, not a marketing line. Vektor exists to make trading understandable for the people it was gatekept from, and the learning stays free.

De-risk by default

Access ships with guardrails attached: deterministic limits, human checkpoints, practice before real money. We think boring guardrails are a feature.

Transparency

It shows its work. Every decision arrives with its reasoning attached, written in plain language and kept on the record.

It explains before it acts

Every signal arrives with its full reasoning chain: what the agent saw, what it weighed, and why it decided. Understanding is part of the product, not a footnote.

You stay in the loop

Today the agent proposes and you decide every trade. That control never goes away. As automation arrives, it will act only inside your limits and confidence threshold, and anything outside the band still comes to you.

Learning comes first

Vektor ships with a free learn hub today: plain-language lessons, a glossary, and quizzes, plus an onboarding step where you set your experience level. In time, lessons will draw on your agent's actual runs: why a card was blocked, what moved a confidence score. Paper mode comes ahead of live trading.

What Vektor does today

One loop: propose, explain, ask

You point the agent at a ticker and the analysis nodes go to work. Supervisor checks market status and your portfolio. Research reads the news and pulls real SEC filings. Strategy forms a signal with a confidence score. Risk applies your limits in deterministic code, and anything that clears the gate lands in front of you as an approval card. Say yes and the risk check runs once more before the order goes anywhere. That is the loop today. Bounded automation will build on it, not around it.

Five-node reasoning chain

Scout scans the universe and picks candidates when the agent runs autonomously. Supervisor, Research, Strategy, and Risk then carry every signal to a decision. The whole chain is recorded, not just the answer.

Approval cards

Signals become cards: the trade, the size, the confidence, and the reasoning behind it. Cards expire rather than linger. Today every trade waits for your approval; in time, only the calls outside your confidence band will need it.

Position watching, automatic exits

Vektor checks every open position continuously and enforces your exit rules for you: stop-loss, take-profit, maximum hold time, and end-of-day close. Deterministic, hard-coded rules, no LLM in the loop.

Deterministic risk gate

The Risk node is hard-coded, deterministic rules that no LLM can override. Position limits and circuit breakers apply identically every time, with a reproducible rule citation.

Honest signals

Strategy pairs LLM reasoning with ML model signals and live market data, and it never bluffs: when a trained model is unavailable, the card says so plainly. No fabricated numbers, ever.

Full audit trail

Every run, signal, and decision is recorded and reviewable. Each night the agent's calls are scored against what the market actually did, and the score sticks to the record.

Live market view

Streaming real-time prices across your watchlist and dashboard, with interactive charts: candlestick or line, from one day out to a full year. Run the agent on any ticker straight from the list.

Why Vektor

Different by design

01

It closes the loop

Most tools stop at a suggestion and leave the follow-through to you. Vektor is built to carry a decision all the way to the order. One system runs from reasoning to execution, and the gate in the middle is yours: your approval today, your limits and thresholds as automation ships.

02

Grounded, not guessing

Vektor is not a chatbot with a ticker feed. LLM reasoning runs alongside deterministic, hard-coded gates, and every judgment is anchored to real signals: news, SEC filings, sentiment scores, ML model outputs, price and volume. And you stay in the loop: today by approving every trade, later by setting the bands the agent works inside.

03

One agent, two markets, regulated rails

Equities and crypto run through a single agent and a single approval loop, built on regulated infrastructure: a FINRA-registered brokerage for equities and licensed digital-asset rails for crypto.

04

A collaborator, not an oracle

Vektor is built to learn from the real outcomes of the trades you approve, and from your feedback on its reasoning. It adapts to how you invest rather than handing down verdicts from a distance.

From signal to decision in four steps

The full loop, from signup to your approval.

01

Sign up, get a brokerage account

The moment you register, Vektor sets up a paper brokerage account for you in the background through Alpaca's Broker API. There is nothing to link and no passwords to hand over. Applying for live trading will be a separate step, with identity checks and Alpaca's own approval.

02

Set your guardrails

Position limits, loss limits, confidence thresholds, approval expiry. These live in deterministic code. The agent works inside them, never around them.

03

Run the agent

Point it at a ticker. You picked the target yourself, so Scout stands down and the four analysis nodes take it from there: Supervisor, Research, Strategy, and Risk, each writing its step into the reasoning chain.

04

Review and approve

Qualifying signals land as approval cards. Read the why, then approve or reject. Before any order is placed, the risk check runs one last time. When bounded automation arrives, this step stays for every call outside your band.

Under the hood

Five nodes, one grounded loop

Three of the five nodes are deterministic, hard-coded rules. Two reason with an LLM over real signals. Click a node to see what it does and what it reads.

Scout

Deterministic

Scores the scan universe with local signals and picks the strongest candidate. It runs on autonomous scans; on a manual run it stands down because you chose the ticker yourself.

Reads

News freshnessPrice momentumVolume vs 20-day averageWatchlist and holdings priorityCached sentiment

Vision · not yet shipped

Where Vektor is going

Everything below is direction, written in the future tense on purpose. When something ships, it moves up this page. Not before.

Flagship

Bounded automation

Vektor will run on a schedule you set, with Scout picking the candidates, and act inside your limits: trades above your confidence threshold will execute on their own, while anything below the band lands in your queue as a card. The deterministic risk gate checks every order either way. The three-band logic already lives in the risk node; what ships next is the schedule you control.

Flagship

Choose the models that power it

You will pick which LLM runs each node: a fast model for research, a deep one for strategy, any mix you like. Your combination shapes the insights you get and sets what each run costs in credits. The pipeline already accepts per-node choices; the picker is what ships next.

Flagship

Approve from your phone

A native iOS and Android app is coming, because that is where your attention already lives. When a call needs you, Vektor will push a notification: open it, read the reasoning, approve or reject. Approval cards already expire today; the app will turn that window into something you can act on in seconds, from anywhere. This is how bounded automation will reach you: acting inside your limits, and pinging you for the edge cases.

An agent you can teach

Your feedback will tune how it explains and what it recommends, always inside your risk limits. It will adapt to you, but switching off the guardrails will never be on the table.

Meets you at your level

Level selection and the learn hub exist today. What comes next is depth that adapts: onboarding will decide whether you see plain-language fundamentals or full trader-grade detail, and the whole experience will grow as you do. Understanding stays free either way.

Learning tuned to your runs

Lessons and explanations will draw on the decisions your agent actually made: why a card was blocked, what moved a confidence score, what to read next. The learn hub ships today; this tailored layer comes next.

Read-only brokerage connect

You will connect an existing brokerage read-only and watch the agent analyze what you already own, before you fund anything.

Paper mode before live

You will run the full loop on simulated money and judge the agent by its record before a real dollar moves.

A public track record

Vektor already scores its calls against real outcomes every night. The next step is making that record public and verifiable, so anyone can check the tape rather than take our word for it.

Smarter approval cards

Calibrated confidence you can act on, plus why-not cards that show the trades the agent chose not to make.

Portfolio-level agency

Drawdown guardrails and tax-loss harvesting. The agent will watch the whole portfolio, not just the next trade.

Simple, honest pricing

Credits, not run counts. Each agent run costs credits based on the models it uses, so heavier thinking costs more and lighter runs cost less. No plan removes your control.

Starter
$0forever

The full loop with you approving each trade.

Coming soon
  • 15 credits per day*
  • Agent runs with the full reasoning chain
  • Approval queue
  • Watchlist with live prices
  • Run history
  • Two-factor authentication
  • Paper mode*
Most popular
Pro
$29per month

More credits, and bounded automation when it ships.

Coming soon
  • Everything in Starter
  • 100 credits per day*
  • Bounded automation inside your limits*
  • LLM selection per node*
  • Price alerts*
Team
$79per month

For teams and high-volume desks.

Coming soon
  • Everything in Pro
  • 400 credits per day*
  • Priority support
  • Early access to new features
  • Team seats*
  • Full US market auto-scan*

* Roadmap features, included in the plan when they ship. Credit budgets are beta values and may be tuned. Beta pricing: changes come with 30 days' notice. No credit card required for Starter. Live trading will require identity verification and Alpaca's approval.

An agent that shows its work and asks first

Vektor is in beta, and early access opens soon. Everything above is the real loop, described exactly as it ships.

Early access coming soon