Daily News 19

Tag: Model Context Protocol

  • Bybit Integrates Chat into AI Stack, Signaling Broader Crypto-AI Shift

    Bybit Integrates Chat into AI Stack, Signaling Broader Crypto-AI Shift

    Crypto exchanges and retail brokers have largely converged on Model Context Protocol (MCP) integrations within their AI roadmaps, but some platforms are now advancing beyond that baseline. Bybit, the second-largest crypto exchange by trading volume, has unveiled Bybit AI, a conversational co-pilot designed for everyday trading and account management support.

    Bybit AI Launches as Conversational Layer Over MCP Infrastructure

    The firm launched an MCP server in April, connecting its platform to AI agents such as ChatGPT and Claude. The new Bybit AI front end sits on top of that infrastructure, aiming to simplify how users interact with the exchange’s suite of financial services.

    “Bybit AI is an important part of our roadmap for the New Financial Platform,” said Ben Zhou, Co-founder and CEO of Bybit. “We want to make it easier for users to access the financial services they need in one place.”

    For now, the co-pilot appears restricted to certain products and services, though Zhou indicated the product will expand to provide a wider range of access.

    “The idea is simple: you tell Bybit AI what you want to do, and it helps you find the right products and services to get it done, like having a team of financial experts right in your pocket,” he said.

    AI Co-Pilots Operate Within Strict Boundaries

    The crypto exchange emphasized that the co-pilot will not replace human agents but will work alongside professionals. Much like other MCP-based implementations, Bybit’s AI co-pilot will not have direct access to a client’s main account.

    Strict security measures to prevent prompt misunderstandings and unintended trades play a key role in these limitations. Additionally, the lack of clear, jurisdiction-specific regulatory rules has pushed brokers and exchanges to move cautiously, enforcing tight boundaries around what AI can execute.

    Industry Trend Toward Chat-Style Interfaces on MCP Rails

    Nonetheless, layering chat-style interfaces and co-pilot tools on top of MCP infrastructure might represent the next phase in the AI-native features race for brokers and exchanges. The fact that MetaQuotes, the leading third-party platform for retail brokerage, launched an integrated AI helper alongside its MCP connection points to this broader industry direction.

  • Etherscan Adapts to AI Agent Internet Takeover With API Rollout Across 60+ EVM Chains

    Etherscan Adapts to AI Agent Internet Takeover With API Rollout Across 60+ EVM Chains

    Etherscan has launched a suite of tools that allows AI agents and coding assistants to access live blockchain data through the explorer’s API.

    The release targets developers building agents that require verified onchain information across more than 60 EVM-compatible networks. Etherscan announced the “Build with AI” launch in a post on X on August 31, 2026.

    What Etherscan’s AI tools include

    The release provides three ways for software to access Etherscan data: a Model Context Protocol (MCP) server, a command-line interface (CLI), and a collection of installable agent “skills.”

    The hosted MCP server can be accessed by agents such as Claude and ChatGPT through natural-language requests. According to Etherscan’s documentation, it runs from a single endpoint at mcp.etherscan.io/mcp and exposes 20 tools.

    These tools cover core blockchain queries, including native and token balances, normal and internal transactions, transaction and receipt lookups, contract information, gas prices, and event logs. A single connection can reach every supported network.

    The CLI follows the structure of Etherscan’s API, using commands in the form etherscan <command>. It returns JSON by default and also supports table and CSV output for users who need results in a more readable format.

    All three components authenticate with a standard Etherscan API key. A free key can be used across every supported chain.

    Skills for tracing funds and reviewing contracts

    Etherscan’s public GitHub repository lists four installable skills. Developers can add them with the npx skills add command or copy the relevant folder into an agent’s skills directory.

    Etherscan Flow traces fund movements between addresses and records them in a case file. Etherscan Contract Review explains the function of a verified, deployed smart contract. A transaction debugger reconstructs what a transaction did and why, while an orchestrator skill directs each task to the appropriate interface.

    According to the repository, the purpose of the skills is that “every address, amount, and transaction hash comes from a live API call, never invented.” The suite is read-only and follows the API’s existing quotas and rate limits.

    Etherscan targets AI hallucinations in crypto

    Etherscan’s release addresses the hallucination problem affecting crypto-focused AI tools. Language models can generate an incorrect wallet balance or describe a transaction that never occurred, potentially causing costly mistakes.

    These errors often arise when models lack access to the necessary data. In many cases, they cannot crawl the web pages where the relevant blockchain information is published.

    Etherscan also warned users not to enter API keys into fake or fraudulent MCP servers claiming to be legitimate. Its documentation states that only one official MCP server exists and warns that listings on MCP marketplaces using the name “Etherscan MCP” are unaffiliated.

    Blockchain data tools for an increasingly automated web

    The growing scale of automated internet activity is increasing demand for tools that can provide reliable data directly to AI agents.

    In early June, AI agents reportedly surpassed humans as the internet’s largest source of traffic. Cloudflare CEO Matthew Prince confirmed the shift and said it happened sooner than he had predicted.

    Cloudflare Radar data later showed agentic bots accounting for 57.4% of web traffic, compared with 42.6% for humans. North America recorded an even greater imbalance, with bots responsible for 68.6% of traffic.

    Humans have since regained the lead, but bots still represented more than 35% of web traffic during the previous four weeks.

    Publishers are also developing ways to make their content easier for AI agents to access. Etherscan’s release follows a similar model by providing machine-readable documentation and a query layer designed for software readers as well as human users.