Claude

Only Google Gemini and Microsoft Copilot run on a search index their parent company crawls itself. Anthropic Claude rents Brave Search , Mistral Le Chat rents Brave too, OpenAI ChatGPT rents Bing plus its own crawler, and Meta AI rents both. The key clue: Claude’s web_search tool exposes a literal BraveSearchParams field, and citation overlap with Brave runs about 86.7%.

Only Five Companies Actually Crawl the Open Web

Before mapping each AI lab to its backend, the key constraint is simple: only five operators crawl the open web at scale. Everything else sold as a “search engine” resells one of those indexes. The five are Google, Microsoft Bing, Yandex, Baidu, and Brave Search, with Mojeek as a much smaller niche sixth.

On February 23, 2026, Anthropic published a blog post titled “How AI Helps Break the Cost Barrier to COBOL Modernization” . It shipped with a Code Modernization Playbook . By market close, IBM’s stock had fallen 13.2% to $223.35 per share. That was IBM’s worst single day since October 2000. More than $31 billion in market cap vanished. Accenture fell 6.5%. Cognizant dropped 6%. One blog post had shaken the whole legacy migration sector.

OpenClaw’s bundled claude-cli backend is officially sanctioned by Anthropic. OAuth-token extraction tools stay blocked. The carve-out works because shelling out to claude -p preserves prompt caching, so a $20 Pro or $200 Max sub routes through OpenClaw without four-figure API bills. The catch used to be a 5-hour usage cap. From June 15, 2026, that claude -p traffic moves onto a separate monthly Agent SDK credit, so the real limit is now a modest dollar budget.

Claude Opus 4.7 landed on April 16, 2026, and after the first 48 hours on X and Reddit the verdict is net-positive but heavily qualified. Power users are calling it state-of-the-art for agentic coding, long refactors, and the viral new Claude Design tool. The loudest complaints cluster around runaway token burn (roughly 1.5-3x more expensive in practice than 4.6), an “ambiguity tax” where the model no longer silently rescues vague prompts, and confidently broken output on marathon runs. Users who prompt like they are writing a spec are getting enormous leverage out of it. Users who prompt the way they used to prompt 4.6 are burning through their usage caps before lunch.

Your AI coding agent has the same file access, shell rights, and database keys you do. A review of 78 studies from January 2026 (arXiv:2601.17548 ) tested every big coding agent. The list ran every major agentic coding assistant . All fell to prompt injection. Adaptive attacks landed more than 85% of the time. This isn’t theory. CVE-2026-23744 gave attackers remote code execution on MCPJam Inspector at CVSS 9.8. A booby-trapped PDF tripped a physical pump through a Claude MCP link at a plant. Attackers hit GitHub’s MCP server to exfiltrate private repository data via malicious issues . And 47 firms fell to a poisoned plugin ecosystem that hid for six months.

The Claude Code skills ecosystem passed 1,340 installable skills in early 2026, and the number keeps climbing. These skills use the universal SKILL.md format : folders of structured instructions that teach AI coding tools to do special tasks. They work across Claude Code, Cursor, Codex CLI, and Gemini CLI without changes. Official skills have shipped from teams at Anthropic, Trail of Bits, Vercel, Stripe, Cloudflare, and dozens of solo devs. Install takes one npx command.