Portfolio Lab Review (2026): Pricing, Features & Honest Verdict

Portfolio Lab is an AI strategy builder that constructs systematic investment strategies, validates them in live paper trading, then lets your Claude or ChatGPT agent trade them in your own brokerage account. Solid pricing, real SEC registration, one badge that contradicts its own disclosures. Best for: self-directed investors. Price: From $20/mo (free plan: yes). Rating: 7/10.

What is Portfolio Lab?

Portfolio Lab sells itself as the strategy layer of agentic trading. You pick an objective, such as steady long-term growth or protection against big losses, and specialised models build a systematic strategy from it: allocations, entry and exit rules, and a cash or hedge position for when conditions turn. The strategy then runs in a live paper account on real prices and fills before a real dollar moves. If you like what it does, your agent trades it in your own brokerage account.

It launched on Product Hunt on 10 August 2026, took 307 upvotes and finished as the number two Product of the Day. The makers credited are Ben Lang and Rich Sun. The product belongs to alphaAI Capital Management LLC, a California limited liability company registered with the SEC as an investment adviser, and the registration is checkable at adviserinfo.sec.gov rather than merely asserted.

The pitch is aimed at a real gap. Systematic strategies, regime models and automated risk management have been institutional tools for decades, and the retail version has mostly been either a robo-adviser that gives you three risk levels, or a backtesting platform that expects you to write Python. Portfolio Lab tries to sit in between. What we found is a company that has thought carefully about the hard parts and got sloppy on one badge in its own hero section.

What are Portfolio Lab’s key features?

One build generates several complete strategies

A single build returns multiple strategies, each tuned to a different objective, and shows you what each holds, when it trades and why. The example on the site runs a tactical long/short over IVV, TQQQ, IBIT, VTV and VHT and returns three variants: one optimised for return at 66.3% CAGR with a 1.24 Sharpe and a 41.1% maximum drawdown, one optimised for Sharpe at 63.8% CAGR and 1.33 Sharpe with a 33.9% drawdown, and one optimised for minimum drawdown at 12.3% CAGR, 0.61 Sharpe and a 12.3% drawdown.

Every one of those figures is labelled simulated and hypothetical, which is the correct labelling. Read the third row before the first. A 66.3% CAGR sitting next to a 41.1% peak-to-trough loss is not a strategy most people can actually hold through, and the site is showing you that trade-off rather than hiding it. Credit where due.

Validation before deployment, not after

This is the part that separates Portfolio Lab from the backtest-and-pray category. Strategies are tested in-sample, validated out of sample, then run in a live paper portfolio against real market prices and fills. The company’s stated method is out-of-sample validation, walk-forward testing across market regimes, point-in-time data and returns net of realistic costs, with the full methodology documented in the app.

Their own framing is the best summary of it: “A misleading backtest is worse than none.” That is an unusually adult thing for a retail investing product to put on its About page.

Agentic trading through your own brokerage

Portfolio Lab does not hold your money. It publishes the day’s plan and your agent, whether Claude, ChatGPT or anything speaking MCP, mirrors that portfolio in your own account and places the orders. The site’s illustrative example shows five orders in a single day: buy VTI, trim NVDA, new AMZN, exit SQQQ, hedge MSFT.

The company is explicit that the language model does not decide anything. It reads, explains and executes, while the allocation comes from the specialised models. That is the right division of labour and worth stating, because plenty of competitors let a chat model pick the stocks.

Named strategy types with published asset limits

Four rulebooks are documented rather than hand-waved. Risk-Aware Hedged runs a long-only core of up to 10 assets and adds a hedge when the models judge risk to be high. Tactical Long-Only holds up to 5 assets and moves to cash defensively. Tactical Long/Short holds up to 5 and shifts between cash and an inverse position. Momentum Long-Only ranks up to 10 assets by momentum and goes to cash when trends turn. The underlying signal set is stated as 7 AI models and 200+ predictors trained on 1B+ datapoints across 35+ years of history and five data categories.

How much does Portfolio Lab cost?

Portfolio Lab starts free and the paid tiers begin at $20/mo. Explorer costs nothing, needs no credit card, and gives you standard ETFs, a live paper portfolio, core metrics and strategy types, and what the site calls “limited weekly builds.” Analyst is $20/mo billed annually or $25/mo monthly, and adds agentic trading, individual stocks alongside ETFs, unlimited builds and deeper analytics. Quant is $40/mo annually or $50/mo monthly, adding leveraged and crypto ETFs, all strategy types and the full metrics suite.

The annual discount is exactly 20% on both paid rungs, $300 down to $240 a year on Analyst and $600 down to $480 on Quant. No games there.

The ladder itself is oddly shaped. Free to Analyst is the jump that changes what the product is: you go from a paper-only sandbox to an agent trading real money, with unlimited builds instead of a weekly cap. Analyst to Quant doubles the price and buys you leveraged and crypto ETFs plus the remaining strategy types. For most self-directed investors that second $20 buys access to the assets they should probably be avoiding. The valuable rung is the first one.

Two numbers are missing. Explorer’s “limited weekly builds” is never quantified anywhere on the site, so the free tier has a cap with no published figure. More significantly, the managed route has no price at all. The homepage offers “Managed by alphaAI. Hands-off, through an SEC-registered adviser” as an alternative to running your own agent, and the disclosures confirm that “Subscription fees are separate from and independent of any advisory fees charged by alphaAI Capital for managed portfolio services.” Those separate disclosures are pointed at alphaai.capital/legal, which returned a 404 when we checked. So if you want the hands-off version, the total cost is not discoverable from either site.

Who is Portfolio Lab best for?

Use Portfolio Lab if you are a self-directed investor who wants systematic rules instead of vibes, and who would rather not build a backtesting stack to get them. The free tier is a real product, not a teaser: you can build strategies and watch them run on live prices in a paper account indefinitely without paying. That is the right way to evaluate something like this, and it costs you nothing but time.

Use it if you already run Claude or ChatGPT with MCP and like the idea of your own agent placing orders in your own account, with the strategy logic coming from somewhere more rigorous than a chat window.

Skip Portfolio Lab if you want someone to manage money for you and tell you the fee up front, because that number is not published. Skip it if headline CAGRs move you more than drawdowns do, because the strategies here can lose 40% of their simulated value and the product is honest enough to show it. And skip it if you are not going to read the disclosures, because the gap between the hero section and the fine print is where the one real problem lives.

What are the best Portfolio Lab alternatives?

Composer is the closest direct rival. Its free plan lets you generate and evaluate trade ideas with AI, build and backtest no-code strategies, and use the API to pull data and trade stocks and ETFs. Trading Pass is $32/mo, or $384/year billed annually for a 20% saving, and adds fully automated investing plus retirement accounts, with a 14-day trial. Composer is more mature at the automation end. Portfolio Lab’s Analyst tier undercuts it by $144 a year and pushes harder on the agentic and validation story.

QuantConnect is the answer if you can write code. The free tier gives equity, index, forex, crypto, futures and options data with unlimited backtesting on one research node and one backtest node with a 500MB workspace. Paid tiers are quote-based rather than listed, which is its own annoyance. You get far more control than Portfolio Lab offers and you pay for it in effort, since nothing is built for you.

Alpaca deserves a mention because it is already underneath Portfolio Lab. The paper trading environment Portfolio Lab uses for validation is Alpaca Securities LLC’s, and Alpaca’s paper accounts and trading API are free to use directly. If you want to write your own logic against the same execution venue and skip the strategy layer entirely, that path exists and costs nothing.

Is Portfolio Lab worth it?

Mostly yes, and the free tier makes the decision cheap. This is a more careful product than the category deserves. The registration is real and verifiable, the backtesting methodology is stated in specifics rather than adjectives, the simulated results are labelled as simulated in every place we looked, and the design decision to keep the language model out of the allocation is the correct one. Paper validation before real money is the feature that should be table stakes across this whole market and almost never is.

The problem is a trust badge. The hero section runs “SEC registered, SIPC member, 128-bit+ encryption” directly above a “Start free” button. The disclosures then say that paper accounts are not real brokerage accounts and are not eligible for SIPC protection, and that SIPC coverage applies only to real-money accounts at Alpaca Securities LLC opened through alphaAI Capital’s managed service. So the badge sits next to the call to action for the one tier it does not cover. For a company this careful about backtest honesty, letting that stand on the homepage is a self-inflicted wound, and it is the single biggest reason this is a 7 and not an 8.

Start on Explorer, build a few strategies, and let them run in paper for a month before you pay anyone anything. If the record holds up, Analyst at $20/mo is fair value for what it adds. Read the disclosures before you read the hero, and get the advisory fee in writing before you go anywhere near the managed option.