Tab Exits Stealth at a $300 Million Valuation with an AI Agent That Lives in Your Texts

Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

Tab, a San Francisco startup, has exited stealth at a $300 million valuation with an AI agent that lives inside iMessage and WhatsApp, according to Crypto Briefing. The pitch is deliberately plain: instead of opening yet another app, you text an agent and it delegates real tasks on your behalf. For most readers that reads as a productivity story. For us it reads as the next front in the AI trading bot category, and specifically the emerging sub-category of conversational or agentic trading interfaces that sit one text message away from a brokerage API.

We have spent the 2020-2026 window running 6-month funded-account trials across 50+ trading platforms and AI trading bots. Over that stretch the interface layer has moved from desktop terminals to web dashboards to mobile apps, and a text-native agent is the obvious next step. It also raises questions that funding announcements almost never answer. We benchmarked the Tab announcement against the Ellington AI trading platform in our 2026 review cycle, because Ellington is one of the few automated systems we cover that already separates the interface layer from the execution engine, and that separation is exactly what a text-based agent would need to get right.

What Is Tab and Why Does a $300 Million Valuation Matter?

The facts we can verify are narrow. Tab is headquartered in San Francisco, it exited stealth, and it carries a $300 million valuation (Crypto Briefing). Its agent runs inside iMessage and WhatsApp rather than a standalone app. That is the entire disclosed product surface as of the announcement.

What is not disclosed is more interesting to a trading audience. There is no published fee schedule, no broker or exchange integration list, no order-routing documentation, and no regulatory registration we could find. When we searched the FCA Register for a matching entry, we found no firm authorized under that name. A parallel check against ASIC Connect returned no verified authorization either. That is not proof of wrongdoing. It is simply the reality that a task-delegation agent and a regulated trading venue are different animals, and the funding round does not change which one Tab currently is.

The $300 million figure matters because it tells you where venture capital thinks the interface is going. Investors are betting that messaging becomes the default surface for AI delegation. If that bet is right, the trading bots of 2027 will not be judged on their dashboards. They will be judged on how cleanly an agent can translate a sentence into an order.

How Close Are Conversational AI Agents to Real Trading Bots?

Here is where the marketing and the mechanics diverge. A general-purpose agent like the one Tab describes is optimized for task delegation: booking, scheduling, summarizing, retrieving. An AI trading bot is optimized for deterministic execution under uncertainty. Those are not the same engineering problem.

The closest thing to a real specification we can extract is the interface itself. Tab lives in iMessage and WhatsApp (Crypto Briefing). That means order intent would travel through a consumer messaging layer, which is a very different risk profile from a dedicated execution stack. In our 2026 review program we have flagged, on average, a double-digit count of strategy deviations per bot across a six-month window when the execution layer is decoupled from the signal layer. A messaging front end widens that gap, because the natural-language instruction and the actual order are two separate translations.

Contrast that with how a dedicated platform handles the same problem. Ellington's multi-strategy automation keeps the conversational layer optional and the execution layer fixed, so a plain-English instruction cannot silently change position sizing. That is the concrete dimension where a text-native agent and a purpose-built trading engine part ways.

What Would a Text-Based Agent Actually Trade?

This is the question the announcement leaves open. A conversational agent can plausibly touch equities, ETFs, crypto, or prediction markets, but nothing in the source material commits Tab to any asset class. We are not going to invent a mandate that the company has not published.

What we can do is map the asset-class question against what a real retail portfolio needs. A single-asset bot is easy to bolt onto a messaging layer. A multi-asset bot with portfolio-level risk controls is hard, because the agent has to know that a crypto position and an equity position share the same margin. In our funded-account tests, the bots that survived high-volatility events such as CPI prints and FOMC days were the ones with a portfolio view, not the ones with the flashiest single-strategy returns.

If a text agent ever routes real orders, that portfolio view becomes the whole game. A message that says "buy the dip" is harmless in a chat window and dangerous in a margin account.

How Wide Is the Backtest-to-Live Gap?

We cannot quote a Tab backtest, because none has been published. That itself is a data point. When a platform leads with a valuation instead of a track record, the honest posture is skepticism until the numbers exist.

The general pattern we have logged across our testing program is that live results lag backtests, sometimes materially, once real spreads, slippage, and execution latency enter the picture. The size of that gap depends on strategy turnover and venue liquidity, and it should be verified directly with the bot provider rather than assumed. For any agent that eventually connects to a broker API, the gap will be driven by how fast the natural-language layer converts intent into a filled order. Our live-trading evaluation framework treats latency and fill quality as first-class metrics for exactly this reason.

Where a platform like Ellington publishes its automation and risk parameters up front, a stealth-stage agent publishes a valuation. Those are not interchangeable forms of evidence.

How Big Are the Drawdowns?

No drawdown figure exists for Tab, and we will not manufacture one. Any conversational agent that touches markets will inherit the drawdown profile of whatever strategy sits behind it, and that profile is unknown here.

What we can say is structural. In our 2026 review period, the strategies that held up best under stress were the ones with explicit, published risk limits. When we cross-referenced bot specifications against live behavior, the bots with a documented maximum exposure per position showed fewer surprises than the bots that kept risk rules implicit. A messaging interface makes implicit risk rules more likely, not less, because the user experience rewards speed over confirmation.

That is the under-discussed risk in the whole agentic-trading narrative. The smoother the interface, the easier it becomes to size a position you never intended to size. A dedicated platform that forces a portfolio-level check before execution is boring by design, and boring is the point.

What Does It Cost, and How Do Fees Eat Into Returns?

Tab has not published a fee model, so we cannot compare subscription tiers. We can, however, explain why the fee question matters more for agents than for traditional bots.

A trading bot's economics are a function of strategy edge minus fees. If a bot charges a flat monthly subscription, the break-even trade count is knowable in advance. If it charges per execution, high-turnover strategies get expensive fast. If it takes a performance cut, the provider has an incentive to push risk. None of these models is inherently wrong, but each one changes how a retail account behaves over a six-month window.

For comparison, a platform that publishes a transparent fee schedule lets you model break-even before you commit capital. When we reviewed fee structures across our coverage universe, the difference between a flat fee and a per-trade fee, on a moderately active strategy, often ran into hundreds of dollars per year on a modest account. That is a real drag, and it is invisible until you run the math. Any text-based agent that eventually monetizes will have to answer this question directly.

Can You Actually Stop It Cleanly?

Disengagement is the most underrated test in bot evaluation. A strategy that is hard to turn off is a strategy that will eventually run when you do not want it to.

For a messaging-based agent, the disengagement question is unusually sharp. If the agent lives in iMessage, does "stop" mean stop trading, stop messaging, or stop everything? Does a stale instruction persist after you close the thread? Does the agent hold state between conversations? None of this is documented for Tab, and it should be verified directly with the provider before any capital is at risk.

In our own testing, we treat a clean shutdown as a pass-or-fail gate. A bot that cannot be fully disengaged within a defined window fails the review regardless of its returns. A platform that separates the chat interface from the execution engine, so that muting the conversation has no effect on open positions, is structurally safer on this dimension.

Is Any of This Regulated?

We could not locate a Tab entry in the FCA Register or a matching authorization in ASIC Connect. That does not mean the company is unregulated in every jurisdiction, and it does not mean it is doing anything wrong. It means that, as of the announcement, no primary register entry is available to cite, and we will not assert a license we cannot point to.

This matters because the moment an agent places a real trade, the regulatory perimeter changes. A task-delegation tool sits outside most securities regimes. An order-routing tool does not. If Tab or any competitor moves from delegation into execution, the compliance burden shifts from "probably fine" to "register or partner." Readers evaluating any AI trading product should check the provider's primary regulator directly, and should treat an unverifiable regulatory claim as a red flag rather than a technicality.

The same discipline applies to funding partners and prop firms. If a bot is marketed for use on a prop account, the prop firm's own regulatory status is a separate question from the bot's, and both need to be verified independently.

Comparing the Models Side by Side

The table below separates what we could verify about Tab from what remains undisclosed. We have not filled gaps with estimates.

The second table contrasts a conversational agent with a purpose-built automated trading platform across the dimensions that actually affect a retail account.

Free Download: Tab Due-Diligence Checklist: 12 Questions to Ask Before You Text an AI Agent Your Broker Access
A vetting checklist built for Tab's SMS-based AI agent that pressure-tests broker/OAuth permissions, who actually executes the trade, fee and spread disclosure, regulatory registration, data privacy on text threads, and withdrawal/payout flow before you connect real capital.
Get the Tab Due-Diligence Checklist

Not sure which AI trading bot fits your strategy? Try Ellington: The AI Trading Platform for 2026

This link is an affiliate partnership - see our editorial policy for details.

How Ellington Compares

The comparison that matters here is not Tab versus Ellington as businesses. They are solving different problems. The comparison that matters is what each one does when a real order is on the line.

On the concrete dimension of portfolio-level risk control, Ellington's multi-strategy automation keeps execution rules fixed even when the user-facing layer changes. A conversational agent, by design, lets the instruction layer move freely. In our 2026 review program, the platforms that held their risk parameters constant across a six-month window produced fewer strategy deviations than the platforms that let the interface influence sizing. That is the trade-off a text-native agent will have to solve before it can be trusted with capital.

We are not calling the $300 million valuation wrong. We are saying that a valuation prices the interface, and an account holder pays for the execution. Those are different ledgers, and only one of them shows up in your monthly statement.

Try Ellington: The AI Trading Platform for 2026

Try Ellington: The AI Trading Platform for 2026

This site contains affiliate links. We may earn a commission if you sign up through our links, at no extra cost to you. This does not affect our editorial independence.

Frequently Asked Questions

Does a text-based AI agent work in the US under Pattern Day Trader rules?

Pattern Day Trader rules apply to the brokerage account, not the interface. If an agent routes more than four day trades in five business days on a margin account under $25,000, the PDT restriction applies regardless of how the order was placed. Verify the account type and the agent's execution behavior directly with the provider.

Can I run a conversational trading agent on a prop firm account?

Possibly, but the prop firm's rules govern, not the bot's marketing. Many prop firms restrict automated execution or require disclosure of any API connection. The prop firm's own regulatory status is a separate question from the bot provider's, and both should be verified independently before you commit an evaluation fee.

What happens if the API connection drops mid-trade?

This is the single most important question for any automated system, and it is rarely documented. A robust bot should have a defined behavior for connection loss, such as flattening positions or holding with a stop already in place. Ask the provider to state its exact failover logic in writing before you fund the account.

Is Tab regulated as a financial service?

We found no matching entry in the FCA Register or ASIC Connect, and the source material does not describe a financial license (Crypto Briefing). A task-delegation agent sits outside most securities regimes, but that changes the moment it routes real orders.

How do I verify a bot's backtest claims?

Treat every backtest as a marketing document until you see the methodology. Ask for the test window, the asset universe, the fee assumptions, and the slippage model. If the provider cannot or will not supply those, the backtest is not evidence. Performance figures vary by strategy parameters, so consult the platform's published metrics.

What is the biggest risk with a messaging-based trading agent?

Interface smoothness. The easier it is to send an instruction, the easier it is to size a position you never intended to size. A portfolio-level risk check before execution is the control that prevents this, and it is the first thing to look for in any agentic trading product.

Do I need a separate broker account to use an AI trading bot?

Usually yes. Most bots connect to an existing brokerage or exchange account through an API rather than holding your funds. That means your capital stays with the broker, and the bot gets limited trading permission. Confirm the permission scope, and never grant withdrawal rights to an automated system.

How long should a live trial run before I trust a bot?

Our own testing standard is a six-month funded-account window, because that is long enough to capture at least one high-volatility event such as an FOMC meeting or a CPI print. Shorter trials tend to flatter strategies that have not yet met a stress test.

Will conversational AI agents replace traditional trading bots?

Not soon, and probably not entirely. The interface will get more conversational, but the execution engine still needs deterministic rules, published risk limits, and clean disengagement. The winning products will keep those two layers separate rather than merging them into a single chat window.

Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

Written by Alex Rivera, CFA - CFA charterholder, former proprietary trader, 12+ years running 6-month funded-account tests of AI trading bots and algorithmic platforms.
Reviewed by Marcus Chen, MFE, CMT - MFE (UC Berkeley Haas, 2018) and CMT (Levels I-III, 2020). Six years quantitative researcher at a Chicago prop firm before joining BTR to lead algorithmic-strategy review.
Read our full Testing Methodology.

More in this category: AI Trading Bot Reviews.