invest
The AI analyst that shows its work.
An AI investment research agent in your terminal. It researches with real data, backtests what it believes, runs a virtual book behind a real risk gate, and grades its own calls, in writing.
$ curl -fsSL https://alo-invest.vercel.app/install.sh | bashFree trial: 3 full agent queries, no card. Bring your own Anthropic API key. $29/mo after — pricing.
Alo Invest is an AI investment research agent for the terminal: it researches companies and markets with live data tools, builds and backtests trading strategies, manages a simulated portfolio behind a mechanical risk gate, and scores its own predictions over time. It cannot place a real trade — by construction, not by promise.
How it works
Watch it think. Then check its math.
01
Ask anything the market can answer
One command, real tools: fundamentals, quotes, filings, news, options chains. The agent reasons out loud, leads with the bear case, and files the call to its journal — so it can be graded later.
> /analyze NVDA ⠿ fundamentals: margins, growth, balance sheet ✓ ⠿ alpaca: quote + 6-month bars ✓ ⠿ perplexity: analyst views, datacenter capex news ✓ Bear case first. Four customers are ~46% of revenue;one capex-digestion quarter compresses the multiple beforeearnings move. Options skew says the crowd agrees on up. VERDICT — structurally long, tactically stretched.Invalidation: hyperscaler capex guidance cut.Filed to prediction journal as call #147 — it will be graded.
02
Build and break strategies
Describe a demand; the factory breeds a cohort, walk-forwards every sibling, and retires the losers. The red-team pass hunts the classic sins — look-ahead, overfit, survivorship — before anything earns a dossier.
> /strategy-factory momentum on liquid tech, weekly rebalance ⠿ exploration map: gap — momentum/equities/weekly untested ✓cohort 2026-07-06: 6 siblings generated ⠿ walk-forward 2019→2026 · 5 folds · cost 5 bps ✓ sibling oos sharpe max dd verdict momo-tech-v1 0.61 -18.2% retire momo-tech-v2 1.12 -11.4% survive momo-vol-scaled 1.38 -9.1% survive …3 more retired RED-TEAM — v2 fills at the close it signals on:look-ahead on rebalance day. Fixed, re-run, then filed.2 of 6 survive the cohort discount. Dossiers written to disk.
03
Run a virtual book with real discipline
Simulated fills at live quotes, behind an actual risk gate. Position caps, cash floors, domain budgets — practice with the walls on. When a trade should be blocked, it is blocked.
> buy $4,000 of QQQ in the book risk gate · 7 checks [PASS] universe: QQQ is approved [PASS] trade_notional: $4,000 within $5,000 limit [PASS] position_size: post-trade QQQ at 4.0% (limit 40%) [PASS] cash_floor: post-trade cash at 46.0% (floor 5%) [PASS] domain_budget: equities 54.0% of NAV (budget 60%)✓ filled 8.6 QQQ @ 463.12 — book NAV $100,412 > now buy $15,000 more [FAIL] trade_notional: $15,000 exceeds max_trade_notional_usd $5,000✗ blocked. The same wall a real book would hit.
Illustrative sessions — the output shapes are real; the numbers are examples.
What it does
Not a chatbot. A research operation.
Six faculties, one discipline: every number from a tool, every claim falsifiable, every call graded after the fact.
- 01Researches
Any company, filing, or macro signal — in seconds.
Fundamentals, news, analyst positioning, options vol regimes, crypto funding. Bear case first, every number from a tool — the agent is not allowed to invent a figure.
- 02Scans
The whole approved universe, mechanically.
Breadth across equities, perps, and options in one pass: funding crowding, earnings where implied move outruns history, vol-regime shifts — surfaced before judgment touches a ticker.
- 03Builds
Strategies like a quant desk, not a chat.
A factory spawns sibling hypotheses against a stated demand, races them through walk-forward and parameter sweeps, retires the losers, and files the survivor with its dossier.
- 04Stress-tests
Its own work, adversarially.
Red-team passes hunt look-ahead bias, overfitting, and survivorship. A promotion gate with a cohort discount — the more siblings you bred, the higher the bar.
- 05Simulates
A virtual book with real guardrails.
Fills computed at live quotes with a cost model — behind the same risk gate a real book faces: universe walls, position caps, domain budgets, a tracer wall on unproven playbooks.
- 06Grades
Itself, against priors it filed first.
Every directional call lands in a prediction journal. Hit rate, Brier score, calibration error — the analyst that cannot mark its own homework generously.
The virtual book
A book that has to argue for its keep.
Every fill at a real quote, every order through the risk gate, every call graded after the fact. Your equity curve becomes the record of an argument — including the drawdowns.
Try it on your hardest question.
3 free agent queries, no card. Then $29/month — cancel any time.
FAQ
Fair questions.
- What is Alo Invest?
- Alo Invest is an AI investment research agent that runs in your terminal. It researches companies and markets with real data tools, builds and backtests trading strategies, manages a simulated portfolio behind a mechanical risk gate, and grades its own predictions over time.
- Can Alo Invest place real trades?
- No. The research edition cannot place real trades by construction: it has no order tools, no brokerage connection, and never holds funds. Its portfolio is a simulation — fills are computed locally from live market quotes.
- Is Alo Invest investment advice?
- No. Alo Invest is a research tool built by Alomana, not a registered investment adviser. Its output is AI-generated analysis for your own research and education, and it can be wrong. Investment decisions are yours; consult a qualified professional before acting.
- What do I need to run it?
- A Mac or Linux terminal and your own Anthropic API key. The one-line installer handles everything else, including Python. No brokerage account, no market-data subscriptions — free data sources work out of the box.
- How is this different from asking ChatGPT about stocks?
- Three ways: every number comes from a live data tool rather than model memory; strategies are validated with real backtests, walk-forwards, and adversarial red-team passes instead of plausible prose; and every directional call is logged and scored — hit rate, Brier score, calibration — so you know when to trust it.
- What is the virtual portfolio?
- A forward-walking simulated book. Orders fill at real market quotes plus a transaction-cost model, and every order passes the same risk gate a real book would face: approved-universe checks, per-trade limits, position caps, cash floors, and per-domain budgets. Oversized trades get blocked, not filled.
- How much does it cost?
- The free trial includes 3 full agent queries with no card required. After that, Alo Invest Research is $29/month through Stripe, cancel any time. You bring your own Anthropic API key and pay Anthropic directly for model usage.
- Is my data private?
- Yes — the agent runs locally. Your prompts, strategies, backtests, and API keys stay on your machine; model calls go directly to Anthropic under your own key. The only thing Alo Invest servers ever see is your license key. No telemetry, no analytics.
Your terminal. Your keys. Your analyst.
$ curl -fsSL https://alo-invest.vercel.app/install.sh | bashNot ready to install? Get the occasional build note instead.