# MCP Articoli collegati

Il Centro Notizie HTX fornisce gli articoli più recenti e le analisi più approfondite su "MCP", coprendo tendenze di mercato, aggiornamenti sui progetti, sviluppi tecnologici e politiche normative nel settore crypto.

OpenClaw Goes Viral, Exposing 12 Types of Critical Vulnerabilities; MCP Protocol Security Benchmark Released

The rapid rise of OpenClaw and similar AI Agents highlights the growing security risks associated with the Model Context Protocol (MCP), a standard enabling models to interact with external tools. Researchers from Beijing University of Posts and Telecommunications introduced MSB (MCP Security Bench), a security benchmark identifying 12 types of attacks across MCP’s three core stages: task planning, tool invocation, and response handling. These include name collision, false errors, retrieval injection, and mixed attacks. Notably, more capable models are often more vulnerable, with an average attack success rate (ASR) of 40.35%. The study also proposes a new metric, Net Resilient Performance (NRP), to balance security and utility. MSB evaluates agents in real-world environments, demonstrating that attacks remain effective even when harmless tools are present. The work underscores the urgent need for robust safety measures as AI agents gain broader tool-use capabilities.

marsbitIeri 04:04

OpenClaw Goes Viral, Exposing 12 Types of Critical Vulnerabilities; MCP Protocol Security Benchmark Released

marsbitIeri 04:04

Whale.io Launches the First AI Agent MCP for Crypto Casino

Whale.io has launched the first AI Agent MCP (Model Context Protocol) for the crypto casino industry, accompanied by a two-week competitive campaign. The MCP enables AI agents to autonomously interact with the platform—placing bets, playing games, and managing real cryptocurrency funds. The campaign features a live leaderboard, performance-based rewards, and a $10,000 USDT prize pool. Aimed at developers and the vibe coding community, it allows participants to deploy AI models—compatible with Claude, OpenAI, LangChain, and others—to test strategies in a real-world crypto casino environment without human intervention.

TheNewsCrypto04/07 19:39

Whale.io Launches the First AI Agent MCP for Crypto Casino

TheNewsCrypto04/07 19:39

60 Curated Skills, Workflows & Open-Source Projects: The Ultimate Claude Advanced List

This comprehensive guide curates 60 essential AI tools, frameworks, and resources to help developers, creators, and builders move from using AI to constructing robust AI systems. The list is organized into nine key categories: AI Programming Agents & IDEs (e.g., Claude Code, Cursor), Agent Frameworks (e.g., OpenClaw, LangGraph), MCP Services & Tool Integrations (e.g., Tavily, Context7), Claude Skills (e.g., PDF Processing, Frontend Design), Local AI & Model Runners (e.g., Ollama, Open WebUI), Workflow & Automation (e.g., n8n, Langflow), Search, Data & RAG (e.g., GPT Researcher, Chroma), API & Infrastructure (e.g., FastAPI, Portkey Gateway), and Learning Resources (e.g., Awesome Agents, PromptingGuide). The author emphasizes a structured approach, recommending specific starting toolkits based on the user's role—developer, creator, product builder, or learner—to avoid overload and build practical, production-ready workflows.

marsbit03/30 11:10

60 Curated Skills, Workflows & Open-Source Projects: The Ultimate Claude Advanced List

marsbit03/30 11:10

AI Agent Economic Infrastructure Research Report (Part 2)

This report analyzes the AI Agent economy, focusing on OpenClaw—a local AI agent that operates autonomously across 20+ platforms like WhatsApp and Slack. It examines OpenClaw's technical architecture, including its message channels, security gateway, ReAct-based reasoning loop, and memory system, highlighting issues like context loss, security risks, and non-deterministic behavior. The study identifies key structural problems in the Agent economy, such as context immobility (locked to local machines) and the "coordination paradox" where multi-agent collaboration lacks trust and verifiability. It argues that crypto infrastructure (e.g., ERC-8004 for identity, x402 for payments) becomes essential only when agents operate across untrusted, cross-platform environments without pre-established trust—enabling micro-payments, decentralized reputation, and auditable logs. While traditional payment giants (e.g., Stripe, Visa) may dominate early adoption, crypto solutions could prevail in the long term due to their superiority in handling high-frequency, cross-border microtransactions and programmable permissions. The report concludes that infrastructure providers (e.g., those offering computation, routing, security) may capture more value than individual agents, and that "Product-Agent Fit" will replace traditional business models, shifting focus to API reliability, data structuring, and chain-verifiable service quality.

marsbit03/24 08:08

AI Agent Economic Infrastructure Research Report (Part 2)

marsbit03/24 08:08

Amid the OpenClaw Craze, CEXs Vie for AI Agent Trading Entry Points

Under the "OpenClaw Craze," major cryptocurrency exchanges like OKX, Binance, Gate, and Bitget are aggressively competing to become the primary transaction entry point for AI Agents. They are no longer just chasing trends but are transforming their core capabilities—trading, market data, wallets, and payments—into modular interfaces that Agents can understand and execute. OKX is focusing on execution, integrating spot, perpetual, and options trading with local signing and simulation modes. Binance is strengthening its information layer and skill distribution, expanding from market data into derivatives and asset management. Bitget is pushing for greater depth, incorporating copy trading, wealth management, and payment services. Gate is building a foundational platform with DEX/CEX MCP, CLI, and Skills Hub. The shift is significant: exchanges are moving beyond AI-assisted analysis to enabling actual system calls and execution. While pre-trade actions like research and screening are becoming smoother, challenges around permissions, security, and risk management in fully automated trading remain. The core competition is about which platform can best modularize its services, ensure security, and become the default financial infrastructure called by Agents in interfaces like Claude and ChatGPT.

Odaily星球日报03/16 06:03

Amid the OpenClaw Craze, CEXs Vie for AI Agent Trading Entry Points

Odaily星球日报03/16 06:03

From Understanding Skill to Learning How to Build Crypto Research Skill

This article explores the evolution and application of Agent Skill, a modular framework introduced by Anthropic in late 2025, which has become a foundational design pattern in the AI Agent ecosystem. Initially a tool to improve Claude's performance on specific tasks, it evolved into an open standard due to high developer adoption. Agent Skill functions like a "dynamic instruction manual" that AI can reference to perform tasks consistently without repetitive user prompting. It is built using a `skill.md` file containing metadata (name and description) and detailed instructions. The system operates through an on-demand loading workflow: the AI first scans lightweight skill metadata, matches the user's intent, then loads only the relevant skill's full instructions, optimizing token usage. Two advanced mechanisms enhance its functionality: - **Reference**: Conditionally loads external documents (e.g., a finance handbook) only when triggered by specific keywords, avoiding unnecessary context consumption. - **Script**: Executes external code (e.g., a Python script) without reading its content, enabling actions like file uploads with zero token cost. The article contrasts Agent Skill with Model Context Protocol (MCP), noting that MCP connects AI to data sources, while Skill defines how to process that data. For advanced use cases like crypto research, combining both is recommended: MCP fetches real-time data (e.g., blockchain info, news APIs), while Skill structures the analysis and output format. A practical example demonstrates building a crypto research agent using an `opennews-mcp` server. The Skill automates workflows like due diligence on new tokens (pulling Twitter data, news sentiment, KOL tracking) and real-time event monitoring (e.g., ZK-proof breakthroughs) to generate structured reports or trading alerts. This combination creates a powerful, automated research system tailored for Web3 analytics.

marsbit03/10 10:41

From Understanding Skill to Learning How to Build Crypto Research Skill

marsbit03/10 10:41

The Silicon-Carbon Co-Governance Journey of a Crypto Company — Cobo's Internal AI Transformation

From its core crypto custody and stablecoin payment operations, Cobo began exploring AI integration in late 2024. Initially, the team experimented with an MCP-based app store but pivoted due to high development costs and lack of standardization. Facing high talent costs and internal resistance, Cobo shifted focus inward, aiming to transform internal operations rather than client-facing products first. A major challenge was security: the company implemented a permission-based internal knowledge system and agent framework using Claude and Gemini with zero-data-retention agreements, ensuring strict data isolation and auditability. Adoption was slow until management enforced AI integration top-down, starting with an OKR Agent that automated goal-setting, progress tracking, and performance reviews. This “silicon-carbon co-governance” approach made AI use mandatory and performance-linked. Over 100 department-specific agents were developed—for customer service, legal, sales, and more—shifting the company’s mindset from hiring more people to deploying AI systems first. Key learnings: healthy cash flow is essential for such transformations; change must be driven from leadership; enforced usage is necessary; and internal AI maturity must precede external AI products. As an outcome, Cobo recently launched WaaS Skill, an AI-agent-integrated financial API layer, reducing development cycles from weeks to conversation-level interactions—a direct result of its internal AI transformation.

marsbit02/25 09:05

The Silicon-Carbon Co-Governance Journey of a Crypto Company — Cobo's Internal AI Transformation

marsbit02/25 09:05

Data Reveals: Is Solana's Slower Transfer Speed Actually Caused by Validators 'Playing Games'?

Jito Labs launched the IBRL Explorer tool to analyze Solana validator behavior in block construction, revealing widespread "timing games" that slow down the network. The tool evaluates validators based on slot time (35%), even distribution of non-vote transactions (40%), and early vote processing (25%). Many validators engage in "late packing," where non-vote transactions are delayed until the final ticks of a slot, prioritizing profit maximization through MEV extraction (e.g., backrunning or sandwich attacks) at the expense of network latency and user experience. This disrupts Solana’s intended streaming design, increases execution variance, and exacerbates negative market structure effects like wider bid-ask spreads. A debate exists between Jito and Temporal (developer of Harmonic client) over what constitutes optimal block construction. Temporal argues IBRL scores favor Jito’s approach and misclassify Harmonic’s auction-based method, which batches transactions but claims continuous execution. Harmonic outperforms in per-block revenue but faces scrutiny over potential user trade-offs. Protocol-level solutions like Multi-Concurrent Proposers (MCP) aim to eliminate single-leader monopolies by enabling parallel block building, but depend on Alpenglow’s mainnet launch (est. 2026). Meanwhile, Jito’s BAM client, now adopted by ~12% of stake, offers auditable ordering logic to mitigate MEV externalities. The competition highlights tensions between validator profitability and network health.

比推01/08 18:21

Data Reveals: Is Solana's Slower Transfer Speed Actually Caused by Validators 'Playing Games'?

比推01/08 18:21

Solana Users Beware: Your SOL Is Being Quietly Harvested in These Ways

A recent article titled "Payment for Order Flow on Solana" has exposed exploitative practices in Solana’s fee market, drawing widespread attention. Similar to traditional finance PFOF models—like Robinhood’s zero-commission trading—Solana applications are leveraging information asymmetry to extract hidden fees from users. Front-end apps and wallets control transaction routing, execution, and fee structures, creating multiple avenues for rent-seeking. These include selling user order flow to market makers, enabling toxic MEV strategies like sandwich attacks, and inflating priority fees and tips. Users—especially retail—are often overcharged due to fear of transaction failure, even when the network isn’t congested. Data shows significant fee disparities: for instance, Axiom users pay median priority fees 200x higher than those paid by high-frequency traders. Much of these excess fees are believed to be captured by the applications themselves, often through kickback arrangements with landing services like Jito. To address these issues, Solana is proposing protocol-level upgrades such as Multiple Concurrent Proposers (MCP) to reduce monopolistic control, Priority Ordering to ensure fair transaction ordering, and a Dynamic Base Fee mechanism to return fee pricing power to the protocol and users. These changes aim to create a more transparent and equitable market structure, essential for Solana’s long-term growth and credibility.

marsbit01/07 06:05

Solana Users Beware: Your SOL Is Being Quietly Harvested in These Ways

marsbit01/07 06:05

Father of Claude Code Reveals: How to Turn Claude into Your 'Virtual Development Team'?

Claude Code's creator, Boris Cherny, shares how to transform Claude from a simple tool into a fully integrated "virtual dev team." Key strategies include: using multiple Claude instances in parallel for different tasks; prioritizing the most capable model (Opus 4.5) for better reasoning; employing a "Plan Mode" to define implementation steps before coding; creating slash commands to automate repetitive workflows; and maintaining a shared team knowledge base to continuously improve Claude's output. He emphasizes setting up automatic verification mechanisms so Claude can self-validate its work, and integrating it with various tools (like MCP) to act as a central orchestrator in the development process. The goal is to make Claude a collaborative, learning partner that reduces errors and manual oversight.

marsbit01/05 00:37

Father of Claude Code Reveals: How to Turn Claude into Your 'Virtual Development Team'?

marsbit01/05 00:37

活动图片

Technology Trends