Bybit AI Skills Launch: 253 API Endpoints Turn the World's Second-Largest Exchange Into an AI Agent Trading Hub
On March 13, 2026, Bybit quietly did something that most exchanges have only talked about: it opened its entire trading infrastructure to AI agents. With a single feature called AI Trading Skill, any major AI assistant — ChatGPT, Claude, Gemini, Cursor, or Windsurf — can now execute trades, pull market data, and manage portfolios on Bybit using plain English. No SDK. No CLI. No configuration files. Just a conversation.
This is not a chatbot bolted onto a trading dashboard. It is 253 API endpoints, organized into six operational modules, designed so that machines can do what humans have been doing manually for years — only faster, around the clock, and without fat-finger errors at 3 AM.
From Human Interfaces to Agent Interfaces
For most of their existence, crypto exchanges have been built for human eyeballs: candlestick charts, order books, buttons labeled "Buy" and "Sell." Programmatic access existed through REST and WebSocket APIs, but using them required developers who could write code, manage authentication, and handle error states.
Bybit's AI Skills flips this model. Instead of requiring users to learn the exchange's language, the exchange now speaks the user's language. A trader can say "set a limit buy for 0.5 BTC at $68,000 with a 2% stop-loss" and the AI translates that intent into the precise API calls, handles authentication, and executes only after the user confirms.
The six modules cover the full surface area of exchange operations:
- Market Intelligence — real-time prices, candlestick data, order book depth, funding rates
- Spot Trading — market and limit orders, batch operations
- Derivatives Trading — leveraged positions, take-profit/stop-loss, conditional orders
- Earn — Flexible Savings and On-Chain Earn products
- Account & Assets — deposits, withdrawals, currency conversion
- Portfolio Management — cross-account queries and position tracking
This is not a toy integration. It is the functional equivalent of giving every AI assistant a Bybit professional trading terminal.
The Security Problem No One Wants to Talk About
Letting an AI execute financial transactions raises an obvious question: what happens when it gets something wrong?
Bybit has embedded several safeguards that deserve scrutiny. New users start on testnet — a simulated environment where AI-driven commands can be tested without risking real funds. When switching to live trading, every transaction requires explicit user confirmation. The AI proposes; the human disposes.
Authentication is handled server-side through the Skill layer, meaning users never need to paste API keys into a chat window or expose credentials to third-party models. Every instruction the AI generates passes through Bybit's platform security checks before execution.
These are sensible defaults. But they also reveal a tension that will define the next phase of AI-driven trading: the more confirmations you require, the less autonomous the agent becomes. At some point, the industry will have to decide whether the value of AI trading lies in assistance (human-in-the-loop) or true autonomy (agent-in-the-loop).
A Three-Way Race for the Agent Economy
Bybit's launch does not exist in a vacuum. It enters a rapidly forming competitive landscape where three distinct approaches to AI-agent infrastructure are emerging simultaneously.
Coinbase: Payments-First with x402
In February 2026, Coinbase launched Agentic Wallets built on its x402 protocol — a payments standard that revives the long-dormant HTTP 402 "Payment Required" status code for machine-to-machine transactions. Wallets can be set up in minutes, support gasless trading on Base (Coinbase's L2), and are designed for a world where AI agents need to pay for services autonomously.
Where Bybit gives agents trading capabilities, Coinbase gives agents wallets. The x402 approach is payments-first: build the rails for machines to move money, and the trading logic will follow.
BNB Chain: Identity-First with ERC-8004
On February 4, 2026, BNB Chain deployed the ERC-8004 standard on mainnet — creating what it calls Trustless Agents. The standard introduces three registries: Identity, Reputation, and Validation. Before an agent can transact, it must prove it is authorized to act.
This is the identity-first approach: solve the trust problem first, then let agents loose on financial infrastructure. ERC-8004 does not compete directly with Bybit's trading-focused AI Skills or Coinbase's payment-focused x402. Instead, it complements both by answering a question neither has fully addressed — how do you know the agent executing a trade is who (or what) it claims to be?
Walbi: No-Code Agent Democratization
Meanwhile, Walbi launched no-code AI trading agents in March 2026 after a 14-week closed beta with over 1,000 participants who created more than 9,500 agents and executed 187,000 autonomous trades. Walbi's approach is arguably the most accessible: traders describe strategies in plain language, and the platform's AI handles execution using portfolio data, technical indicators, the Fear & Greed Index, and macroeconomic signals.
With 2.9 million registered users and a forthcoming agent marketplace where experienced traders can share strategies (complete with transparent performance data), Walbi targets the retail segment that Bybit's more technical integration may not reach.
The Numbers Behind the Shift
The scale of AI-driven trading is already staggering. Analysts estimate that 60–80% of global crypto trading volume is AI-driven today, with projections approaching 90% in the near term. At the 2026 Silicon Valley AI x Crypto Expo, one fund reported that implementing AI agents dropped trading response times to milliseconds and delivered annualized yields 12.3% higher than human-led teams.
The AI agents crypto market cap sits at approximately $3.06 billion as of March 2026 — still small relative to overall crypto, but growing rapidly as infrastructure matures. AI quant funds reported an average return of 52% in 2025, while 84% of retail traders lost money over the same period. The performance gap is becoming the strongest argument for agent-assisted (if not agent-led) trading.
What This Means for Exchange Architecture
Bybit's move signals a deeper architectural shift. Exchanges are no longer just marketplaces — they are becoming API-first platforms where the primary "user" may be a language model, not a person.
This has several implications:
Latency becomes conversational. Traditional exchange APIs optimize for milliseconds. AI agent APIs must optimize for natural language parsing, intent extraction, and confirmation flows. The bottleneck shifts from network speed to comprehension accuracy.
Documentation becomes the product. When your users are AI models, the quality of your API documentation, error messages, and response schemas directly determines whether agents can use your platform effectively. Bybit's 253 endpoints must be machine-readable in a way that goes beyond typical developer documentation.
Compliance gets complicated. If an AI agent executes a trade, who is responsible — the user who issued the natural language command, the AI that interpreted it, or the exchange that executed it? Current regulatory frameworks do not have clear answers, and the SEC-CFTC Joint Harmonization Initiative launched in March 2026 will eventually have to address agent-mediated trading.
The moat becomes the ecosystem. Bybit's advantage is not just 253 endpoints — it is the growing ecosystem of AI assistants that can connect to those endpoints. As more models integrate Bybit's trading capabilities, the switching cost for users increases with each AI assistant they configure.
The Road Ahead: From Assistants to Autonomous Agents
Today, Bybit's AI Skills operates in what the industry calls "co-pilot mode" — the AI suggests, the human approves. But the technology trajectory points toward increasing autonomy. The question is not whether fully autonomous AI trading agents will emerge, but when exchanges will be comfortable enabling them.
The building blocks are falling into place: Bybit provides the trading infrastructure, Coinbase provides the payment rails, BNB Chain provides the identity layer, and platforms like Walbi provide the no-code interface for retail adoption. What is missing is the regulatory framework that defines liability, the insurance products that cover AI trading losses, and the audit standards that verify agent behavior.
For now, Bybit's launch is the clearest signal yet that the world's largest exchanges see their future not in building better dashboards for humans, but in building better APIs for machines. The 253 endpoints are just the beginning. The real question is what happens when that number reaches 2,530 — and the agents no longer need to ask permission.
For developers building AI-powered trading tools and blockchain applications, robust infrastructure is the foundation everything else depends on. BlockEden.xyz provides enterprise-grade RPC endpoints and API services across major chains including Ethereum, Sui, Aptos, and more — the kind of reliable backend that both human developers and AI agents need. Explore our API marketplace to build on infrastructure designed for the agentic era.