DataDome vs WorkOS

Introducing DataDome’s Hosted MCP Server for Direct AI Agent Access to Trend Reports

DataDome announced the launch of its Hosted MCP Server, enabling direct AI agent access to trend reports. This product update expands DataDome's AI agent trust and management capabilities.

Scraping Sites Protected by Cloudflare, DataDome, PerimeterX

This technical guide explains how modern anti-bot systems block web scrapers and describes practical, probabilistic strategies to collect public data reliably. It outlines four independent detection layers—IP reputation, TLS/HTTP fingerprint, a JavaScript sensor, and behavioral signals—and explains why simple header spoofing fails. The article compares vendor behaviours (Cloudflare, DataDome, PerimeterX/HUMAN, Akamai, Kasada), shows how clearance cookies are IP-bound, and recommends an escalation pattern: Chrome-impersonated HTTP, hardened stealth browsers, and racing fresh IPs with cookie reuse. The guide also contrasts IP tiers (datacenter, residential, mobile), warns that success is never 100% and stresses counting only real pages as successes. It positions Crawlora's Web Scraping API as an example service implementing these techniques.

• Modern anti-bot systems evaluate four independent layers: IP reputation, TLS/HTTP fingerprint, a JavaScript sensor, and behaviour over time.
• Cloudflare issues a cf_clearance cookie bound to IP and User-Agent after a managed challenge; changing IP voids the cookie.
• DataDome scores requests in real time, sets a datadome cookie, and is aggressive about datacenter IP ranges and fingerprint replay.

AI visibility shifts from referrals to agentic distribution

Publishers are shifting their focus from expecting referral traffic from AI answer engines toward treating AI agents as a distribution layer they must control and monetize. Multiple industry reports show rapid growth in agentic AI traffic (DataDome, Decodo/Cloudflare) and rising adoption of agent-readable web standards like LLMs.txt (Originality.ai), but usage remains low. Webflow analysis finds median sites appear in a minority of AI answers and receive few citation links. Many publishers are blocking or whitelisting bots (HasData; Reuters and Time examples), but technical limits mean blocking is imperfect. Industry voices urge publishers to build nuanced crawling, indexing, and monetization policies based on agent identity, purpose, and business value rather than blanket allow/block rules.

• DataDome reported 17.7 billion AI agent requests in April–June 2026, a 45% increase from Q1 2026 (12.2 billion); June 2026 alone had 6.6 billion requests.
• Meta generated the majority of AI agent traffic on DataDome’s network; Meta’s training crawler grew 74% Q1→Q2 2026 and its RAG crawler grew 163% in the same period.
• A July Decodo report analyzing Cloudflare data found AI-driven traffic grew ~187% in 2025 and that automated systems generated 57.4% of web requests versus 42.6% from humans.

How Two XAI Designers Use Grok Bot for Their Jobs

In this episode of the 'How I AI' podcast, host Claire Vo interviews John Bai and Peng Zheng, designers on the Grok Bot team at SpaceX AI (xAI), about their use of AI agents in daily workflows. Peng demonstrates a check-in pipeline where he sends photos or location names to a custom Grok Bot, which uses Google Places API and image generation to create 3D miniature visuals for his personal website, automating publishing. John shows how he uses Grok Bot with Figma MCP to edit designs, create marketing materials from templates, and generate interactive prototypes from voice commands. They discuss how AI reduces tedious work, enables creative exploration, and shifts designers toward higher-level decisions, emphasizing configuring bots with clear, narrow responsibilities for tasks like email triage and calendar management. The episode is sponsored by WorkOS and Vanta.

• John Bai and Peng Zheng are designers on the Grok Bot team at SpaceX AI (xAI) and appear on the 'How I AI' podcast.
• Peng Zheng built a self-updating personal website with a check-in pipeline using Grok Bot and Google Places API to generate 3D miniature visuals and automate publishing.
• John Bai uses Grok Bot with Figma MCP to edit designs, create marketing materials, and generate interactive prototypes from voice commands.

Grok Bot Built in a Month by Small Team

Roman Ugarte, who led Growth at Cursor, details how he and a small team built Grok Bot for SpaceXAI from scratch in four weeks, then launched it publicly three weeks later. The article covers key decisions, such as building independently rather than integrating with Cursor, and the team's manual onboarding of nearly 300 initial users. Early product choices and a 'colleague-pilled' philosophy contributed to the bot's success. Roman also discusses moats and Cursor's competitive strategy.

• Grok Bot was built from scratch in four weeks by a small team at SpaceXAI.
• Roman Ugarte previously led Growth at Cursor, scaling it from 15 to over 1,000 employees before acquisition by SpaceX.
• The team manually onboarded nearly 300 of the first users.

Build an AI Code-Review Agent in 30 Minutes

A Lenny’s How I AI episode showcases two use cases for modern LLM-powered agents: Claire demonstrates building 'Merge Mommy', an AI GitHub agent that reviews pull requests, scores their risk across six dimensions, auto-approves low-risk PRs, and routes questionable ones to Slack — all built in a single Codex session and deployed with Vercel Eve. Grace Clarke describes using Claude Code to run three reusable business skills (pipeline, proposal builder, voice guide), replacing Gmail with a Claude-powered inbox and emphasizing intent engineering, skill files, and habitual use over perfect prompts. The piece highlights practical operational controls (risk thresholds, audit logs, SOC 2 alignment) and argues that infrastructure like Vercel Eve reduces setup friction for internal agents.

• Claire built an AI agent that reviews pull requests, scores risk, auto-approves the safest ones, and sends questionable PRs to Slack.
• Intercom reported that PRs approved by its AI system move five times faster than human-reviewed PRs and have a lower revert rate (as cited in the episode).
• The described risk model scores PRs across six dimensions (change size, blast radius, reversibility, data/security implications, operational impact, tests/CI); <24 points is low-risk (auto-cleared) and >64 points goes to humans.