Monthly edition · November 2026 · frozen Nov 9, 2026

State of AI trading agents: November 2026

Name: State of AI trading agents: November 2026
Creator: TradingAgentStack
Published: 2026-11-09
License: https://creativecommons.org/licenses/by/4.0/

4 of the 461,747 ERC-8004 agent registrations we have indexed so far have a trading record we could verify from venue data, and 17 of 18 advertised AI trading returns can't be verified from outside the company. This edition records what could be measured on Nov 9, 2026. It is frozen, so a number you quote from it will still be here.

Cite this edition Download the numbers (JSON)

Each number links to the page that shows how it was computed. For today's numbers see the statistics hub, which updates daily. Past results don't predict future results. Not financial advice.

4 of 461,747indexed ERC-8004 agents with a trading record we could verify

17 of 18advertised returns nobody outside the company can check

3 of 5AI-labelled Hyperliquid vaults beat BTC over the same dates

6 of 12broker AI connectors place real orders without asking first

How much of AI trading can anyone verify?

Most agents and bots publish no record that an outsider can check. These numbers count how many do.

4 of 461,747
ERC-8004 agent registrations we have indexed so far have a trading record we could compute from venue data. 26,254 have a registration file we could read, and 10,489 say they trade or have a wallet active on Hyperliquid. Indexed in full: Ethereum and Arbitrum. Still being indexed: Base and BNB Chain.ERC-8004 agents that trade · as of Nov 9, 2026
0
validations are recorded for ERC-8004 agents, from 0 validators, according to the 8004scan explorer. It counts 2,359,461 registered agents across the chains it covers and 4,761,569 feedback entries with an average score of 98.4.8004scan · as of Nov 9, 2026
17 of 18
advertised AI trading returns we checked can't be verified by anyone outside the company: 11 are claims with no public record and 6 are backtests or simulations. 1 we measured ourselves.Claim check

What the verifiable ones did

Computed by us from public chain and venue data. These are the only results anyone can check.

0.0%
median 30-day return of the 17 vaults and pools that pass both gates (venue Safety score of 50% or more, never down more than 30% from a peak), out of 29 we track. BTC over the same 30 days: −10.6%. 14 of 17 beat BTC.Weekly report · TAS Agent Index method · as of Nov 9, 2026
3 of 5
Hyperliquid vaults whose leaders say they're AI-run beat BTC over the same dates. 1 lost money, and 0 fell more than 30% from a peak.Leaderboard: AI-labelled vaults · as of Nov 9, 2026
136 of 162
AI agents with enough activity on prediction markets made money over 90 days. Across all 310 active agents, trading ROI was +8.4% before AI-tool costs. 3,887 AI trader agents have registered in the Olas service registry.AI prediction-market agents · as of Nov 9, 2026
2 of 9
AI trading apps on Hyperliquid had users with a net realized loss over 30 days. Together their 1,493 users traded $396.4M, and 44% of users who closed a trade were in profit.AI trading apps on Hyperliquid · Oct 10, 2026 to Nov 8, 2026
31 of 38
AI model accounts lost money in Alpha Arena's two real-money seasons. 11.3% of the starting money went to trading fees.Alpha Arena recomputed · Oct 17 to Dec 3, 2025

What can go wrong

Results don't show what a bot or agent can do with your money. These do.

6 of 12
official broker AI connectors (MCP servers) let an AI agent place real orders without asking you first.How broker MCPs gate orders
20 of 109
tools that can trade or move money have a Safety score below 40 out of 100, the band with the most money at risk if they, or an agent using them, are wrong, hacked or prompt-injected.Safety scores
20
security incidents and security releases on trading bots, exchanges, wallets and agent tools dated in the 30 days to Nov 9, 2026, each from an official disclosure. The tracker holds 23, the earliest from Sep 10, 2026.Incident tracker
5 of 45
open-source trading and agent repos we list have published security advisories, usually with a fix.Open-source repo health

Where the money goes

Fees measure demand, not whether users made money. Most of these products publish no trading record.

$41.7M
in fees users paid 28 trading apps and terminals in 30 days, led by Axiom ($21.1M).Bot usage stats · DeFiLlama · as of Nov 9, 2026
$4.7M
in fees users paid 15 Telegram trading bots in 30 days; 76% went to one bot (Terminal).Bot usage stats · DeFiLlama · as of Nov 9, 2026
$1.8M
in fees in 30 days across the 23 AI agent protocols that report fees or deposits on DeFiLlama, with $21.5M deposited.Bot usage stats · DeFiLlama · as of Nov 9, 2026

From official announcements

Events that change what can be verified, each from the organization's own page.

Oct 30
The Recall Foundation, which ran public trading competitions and rankings for AI agents, announced that it is winding down: new competitions and staking rewards stopped, and its app becomes read-only on November 30, 2026.Recall Foundation

Change since the last edition

This is the first edition, so there is nothing to compare it with yet. From the next one, this section shows each number next to last month's.

Cite this edition

Free to use under CC BY 4.0. Please credit and link the edition, which doesn't change: TradingAgentStack, "State of AI trading agents: November 2026", https://tradingagentstack.com/state-of-ai-trading-agents/2026-11/

The numbers as one file: state-of-ai-trading-agents-2026-11.json. The records behind them are in our open data, and agents can query them through the MCP server. How each number is made: how we verify results and how we score risk.

Writing about AI trading agents? If you need a number that isn't here, the method behind one, or the raw records, write to hello@tradingagentstack.com. We answer with the data and how it was computed.

All editions

November 2026 this pageFrozen Nov 9, 2026 · 4 of 461,747 indexed ERC-8004 agents verifiable