Claude / Anthropic
News, releases, and guides for Claude and the Anthropic developer ecosystem.
bypassPermissions drops every gate — use these modes instead
acceptEdits, auto, dontAsk, and allow rules cut friction without the full trust loss of bypassPermissions.
OpenAI's Codex plugin hands your code review to a second LLM
OpenAI's Codex plugin runs inside Claude Code to delegate code review, adversarial critique, and tasks to Codex locally.
free LLM API resources GitHub repo list free providers rate limits Claude Code Cursor 2026
You don't need a paid API key to run a coding agent anymore — a small category of GitHub directories now catalogs every inference provider with a genuinely free tier, complete with rate limits and drop-in configs. The seed for this piece is a walkthrough video, but the
OpenAI ships Codex into Claude Code — two commands, or four?
OpenAI's codex-plugin-cc adds Codex review to Claude Code. The real install path is four commands, not two.
MengTo/Skills has no tagged releases — pin the commit
78 MIT agent skills across web design, GSAP, Three.js, Codex. No releases in MengTo/Skills — pin a commit.
36 languages parsed locally — Graphify needs no model call
Graphify indexes your codebase with tree-sitter across 36 languages — local-only, no embeddings, no API calls.
Anthropic's Finance plugin: README says 6 skills, folder has 8
Finance and Legal Claude plugins: skill dirs, slash commands, MCP connector categories, and README-vs-folder gaps.
OmniRoute's "unlimited free" claim — 90 providers, not 200
OmniRoute gives Claude Code 90+ free-tier providers via quota fallback — not the 200+ free that viral coverage claims.
One Claude session, 20 front-ends — that's Hermes
Hermes: one Claude session across Telegram, Slack, and 20+ apps. Install, provider config, and trigger setup covered.
No runnable check? You become Claude Code's verification loop
Four ranked skills from Anthropic's Claude Code docs: verification, plan mode, context hygiene, CLAUDE.md.
The env var that silently routes Fable agents to Haiku
How CLAUDE_CODE_SUBAGENT_MODEL precedence works in Claude Code, and why it silently overrides Fable 5 pinning.
One SKILL.md replaces CLAUDE.md, AGENTS.md, and .cursorrules
Agent Skills open standard: SKILL.md replaces CLAUDE.md, AGENTS.md, and .cursorrules for Claude Code, Codex, and Cursor.
Claude Code artifacts publish a live URL — the chat sandbox
Claude Code artifacts vs chat artifacts: supported types (.html/.htm/.md), CSP sandbox limits, MCP connector support
Anthropic blocked OAuth in third-party tools — here's what
Anthropic cut off Pro/Max OAuth tokens from OpenCode, Cline, RooCode in April 2026. API keys and Bedrock/Vertex still
Claude Code subagents concurrent parallel architecture context isolation limits 2026
A single operator running a "department" of AI employees — one researching, one writing, one analyzing — is the pitch behind a viral demo, but Claude Code's real parallelization story is more precise, and more limited, than the headline suggests. How Claude Code's Parallelization Architecture Evolved
Claude Code skill files load in ~100 tokens — then pull the
How SKILL.md's three-tier progressive loading turns UI design extraction into a reusable, version-controlled operator
20+ agent backends, one Artifacts canvas — self-hosted
Apache-2.0, local-first canvas supporting 20+ coding agents as a self-hosted Claude Artifacts alternative.
launchd KeepAlive + tmux detach = respawn loop — the fix
Why launchd KeepAlive + detached tmux respawn-loops Claude Code and the safe RunAtLoad bootstrap pattern to avoid it.
Three Claude spend trackers, one 17k-star engine underneath
All three macOS Claude Code cost trackers shell to ccusage — differences are UI, install path, and rotation health.
Claude Code automation primitives loops goals routines dynamic workflows
Anthropic has repackaged automation in Claude Code as four composable primitives — loops, goals, routines, and dynamic workflows — that turn the developer from "the loop" into the designer of a loop that runs itself. Each one wraps the same turn-based cycle every Claude Code prompt already executes: evaluate
Claude Code hooks silently fail on Hermes — guards don't fire
The agentskills.io spec, Claude Code's hook extensions, and which skills actually port to Hermes Agent's open runtime.
One user's $35K tab is why Claude Max has a weekly cap now
Claude Max weekly caps: 5-hour window, 7-day pool reset, overflow billing, and why parallel agents drain the limit fast.
Ponytail cuts Claude output 54% — what the benchmark measures
Ponytail adds a YAGNI ladder to Claude Code: benchmark shows −54% lines, −22% tokens, 100% safety pass rate.
video-use feeds Claude Code a transcript, not 45M frame-tokens
video-use (browser-use org) edits video via packed transcript and ffmpeg EDL pipeline — no timeline UI required.
What's inside a Chrome extension that ports AI chat context
MV3 extensions bridge AI chatbot sessions via DOM scraping — compared on privacy, cost, and automation.
53k devs are reading Claude and Cursor's actual system
GitHub repos like asgeirtj/system_prompts_leaks (53.5k stars) and x1xhlol (139k+) archive extracted system prompts for
Vercel's skills.sh shows 900K+ — the curated catalog is 342
skills.sh all-time: 900K+. Curated API: 342 skills across 87 owners. What each number counts and why the gap exists.
SkillUI clones a site's design system — no AI, no API key
SkillUI v1.3.4 extracts colors, typography, and spacing from any URL into a Claude Code SKILL.md — no AI, no API key.
What MotionSites actually ships (it's not a website builder)
MotionSites AI sells scroll prompts for Claude Code — no runtime. How GSAP output compares to Framer AI and v0 in 2026.
600 skill improvements, one background hook — task-observer
Background hook parses JSONL transcripts, logs gaps, and queues SKILL.md rewrites — 600+ edits across 40 skills.
Sonnet 5 is default — your token count just jumped ~30%
Claude Code v2.1.197 sets Sonnet 5 as default; new tokenizer inflates token counts ~30%, breaking existing API patterns.
Claude's MCP OAuth locks to claude.ai — n8n routes around it
Build n8n workflow tools that bypass Claude's OAuth redirect lock, 25k token ceiling, and unsupported MCP connectors.
n8n turns your workflow into an MCP server — stdio never
How n8n's MCP Server Trigger exposes n8n workflows as MCP tools over SSE/Streamable HTTP, letting Claude invoke custom
Without a verification gate, Claude Code quits at 'looks done'
Verification gates, /compact flags, terse CLAUDE.md, and fan-out subagents — Claude Code practices most developers skip.
You're using Claude Code with the parking brake on
32 Claude Code patterns: context management, CLAUDE.md, parallel worktrees, lifecycle hooks, ultrathink trigger.
What practitioners encode in SKILL.md — and why it now travels
Claude Code SKILL.md: progressive disclosure, five practitioner competencies, scope hierarchy, and common failure modes.
A Bloated CLAUDE.md Teaches Claude to Ignore You
CLAUDE.md pruning, PreToolUse hooks, subagents, and headless CI for Claude Code power users. Updated through v2.1.198.
Shell failures now explain themselves in Claude Code
Claude Code weeks 24–26 in June 2026: /cd, recursive delegation, live pages, shell auto-response, and Sonnet 5.
Your linter reports. /simplify did the cleanup for you.
Three parallel agents, one slash command: how /simplify auto-applies diff cleanup before your PR is opened.
/btw, /compact, /rewind: what most Claude Code users ignore
The Claude Code features most teams skip: /btw, /compact, /rewind, Stop hooks, and CLAUDE.md import patterns.
A 5× expertise gap separates novice from proficient AI pairing
398k Claude Code sessions: occupation outpredicts programming skill, expertise multiplies return 5×. Caveats apply.
58 MCP tools fill Claude's context — n8n can reduce that to 4
Turn n8n into a Claude MCP server with curated tools — reducing definition bloat and tool-count overload in minutes.
Instructions Claude follows. Hooks it can't ignore.
Advanced Claude Code techniques: hooks, subagents, skills, plan mode, and context compaction — as of v2.1.145+ (2026).
Your CLAUDE.md is probably too long to obey
CLAUDE.md hierarchy, plan mode, slash definitions, and sub-agent delegation — Claude Code official best practices for mid-2026.
Korea built on Claude — then the U.S. export order hit
Seoul office open June 17. API unchanged for Korean builders. A U.S. export order cut off top-tier models for 15 days.
n8n MCP hub: tool sprawl shrinks, Claude limits don't
Wire n8n's MCP Server Trigger to Claude Desktop: what it offloads, what it doesn't, and where SSE connections break.
CLAUDE.md advises. Hooks overrule. Know which is which.
Official Claude Code best practices for 2026: CLAUDE.md, hooks, slash commands, subagents, and /rewind explained.
Claude activations, now legible. Fidelity: 0.6–0.8.
Anthropic's NLA: verbalizer/reconstructor loop, 0.6–0.8 FVE, Claude test-awareness, and open checkpoints for Qwen2.5, Gemma-3, Llama-3.3.
CLAUDE.md bloat silently causes Claude Code to improvise
Mid-2026 Claude Code guidance: CLAUDE.md sizing, hooks, subagents, MCP, and slash commands for agentic efficiency.
Opus 4.8 and the 11-day Bun port: what the vendor claim hides
Opus 4.8 flags code flaws 4x more, deprecates `budget_tokens`, and adds Dynamic Workflows for long-running pipelines.
The world's top Claude user is Korean — Seoul office follows
Anthropic's Seoul office opened June 2026 — NAVER, Samsung, MSIT safety MOU, and Claude for Startups now live in Korea.
Snowflake's CEO declared a tie. The iteration ledger didn't.
Snowflake CEO tested GLM-5.2 vs Claude Opus 4.7: 103 tasks, pass@3 near-tie, 2x token use, 5.7x cheaper output. Here's what the numbers actually mean for builders.
Ten patches blocked git wipeouts. The annotated sprint.
Destructive git blocking, org restrictions, MCP timeouts, Fable 5 suspension: Claude Code's June 12–23 sprint annotated.
An accountant can outperform a senior dev on Claude Code
Domain expertise, not coding background, predicts Claude Code success. Key data: 398k sessions, Oct 2025–Apr 2026, published June 16, 2026.
Novices quit. Experts adapt. 400k Claude Code sessions say so.
Anthropic's 398k-session study finds Claude Code widens expertise advantage — 15% novice success vs. 33% for experts.
Fugu beats each component it calls — and the asterisks matter.
Fugu coordinates frontier LLMs via one API. What the benchmark sheet covers, what it omits, and caveats for builders.
AlphaFold's creator moved to Anthropic. The IP did not.
AlphaFold co-creator John Jumper (Nobel 2024) is leaving DeepMind for Anthropic. His new role remains unconfirmed.
Blackmail dropped from 96% to 0%. Here's the asterisk.
May 2026 alignment paper: how Anthropic cut Claude's blackmail rate from 96% to 0% and what the limits are.
Activations into English: 4× better at surfacing hidden goals
Anthropic's NLAs map activations to English, exposing hidden goals 4× more than SAEs — and where they confabulate.
Opus 4.8 is a one-line swap. The xhigh recalibration isn't.
Opus 4.8 vs 4.7: +4.9 pts SWE-bench Pro, xhigh recalibrated, GA subagent fleets. Drop-in API; effort tiers changed.
Anthropic writes the checks. CodePath signs the W-2s.
$150M fund, 1,000 fellows embedded at nonprofits, W-2 from CodePath — what Claude Corps covers and who qualifies.
Every Opus 4.8 chart beats 4.7. The asterisks matter.
Opus 4.8 vs 4.7: SWE-Bench Pro up 4.9 pts, 1M-context recall nearly doubled, Dynamic Workflows launched, pricing flat.
27-year-old OpenBSD flaw. Glasswing's in 15 countries now.
Glasswing expanded to ~200 orgs in 15 countries. Mythos Preview found 10,000+ flaws; under 1% are remediated.
K2.7 deliberates on every call — the off-switch doesn't exist
K2.7 Code forces reasoning on every call, caps replies at 32K tokens, and ships under a Modified MIT license — here's the full deployment picture.
The jailbreak wasn't universal. The shutdown was.
US export order suspended Fable 5 and Mythos 5 June 12. What the jailbreak was, who was cut off, and what to migrate to.
v2.1.172 made nesting 5 deep possible. v2.1.181 had to cap it.
Claude Code v2.1.170–v2.1.181, June 9–17: Fable 5, scoped permissions, sub-agent nesting, /config inline annotated.
Refusals now return HTTP OK — and your try-except misses them
v0.108–v0.109 adds refusal responses at HTTP 200, a new stop_reason literal, and removes retired model strings from typed enums.
v0.109: Fable 5's refusal arrives as HTTP 200
SDK v0.108–0.109: Fable 5 refusal-in-200, Managed Agents cron, vault credentials, and an enum fix in 0.109.1.
LLM blackmail: 84% to 0% — and 'why' beat 'what' by 28×
SDF, synthetic stories, and 3M moral-dilemma tokens: how Anthropic cut LLM blackmail from 84% to 0% in May 2026.
$85k, no CS degree — is Claude Corps goodwill or a sales play?
$150M, $85k/year, no CS degree: what Claude Corps offers nonprofits and fellows, plus the critique it's earning.
Anthropic paused the SDK billing split on June 15 itself
Anthropic paused the SDK credit split. Live usage credits system, CI cost baselines, and what triggers a restart.
Opus 4 retired. Opus 4.8 costs 67% less — mind the tokenizer
Sonnet 4 and Opus 4 retired June 15 — exact model IDs, breaking changes, and the Opus 4.8 tokenizer caveat explained.
Fable 5 is export-controlled. So is your colleague's session.
Fable 5 and Mythos 5 export directive: deemed export scope, Glasswing fallout, and licensing conditions.
Fable 5 refusals are 200 OK — your error handler misses them
Fable 5 and Mythos 5 are offline under a U.S. export-control order. Wire up refusal handling and Opus 4.8 fallback now.
MAI-Thinking-1 beats Anthropic's top model — per Microsoft
Seven MAI models at Build 2026. MAI-Thinking-1 is a 35B-active sparse MoE — specs, claimed scores, and what's still unverified externally.
Opus 4.8 kills budget_tokens — here's what else moved
Opus 4.8: fast mode, mid-session system prompts, 1K cache floor. Old budget_tokens syntax returns 400.
Opus 4.8 Thinking Blocks Were Silently Corrupting on Retry
Thinking blocks on Opus 4.8 were corrupting on retry. v2.1.156 is the hotfix — update, verify, and see what else landed.
Your Claude Code Skills Now Hot-Reload Without Restart
Claude Code v2.1.157 adds .claude/skills/ live-loading, worktree unlocking, and OTEL telemetry. Annotated guide.
Claude Code Now Gates Execution on Bedrock and Vertex
v2.1.158 enables classifier-gated execution on managed inference platforms. Here's the env var, what it does, and what to verify before upgrading.
Devin's $26B Raise Changed How AI Coding Agents Compare
From GitHub Agent HQ to Devin's $26B raise: a technical breakdown of what changed in AI coding agents in 2026.
Inject Constraints Mid-Run Without Breaking the Prompt Cache
Mid-conversation constraint injection in v0.105.0 preserves prompt cache continuity across long inference runs.
Anthropic SDK 0.105.0 Needed Two Hotfixes — What to Pin
Two rapid patches followed Anthropic's 0.105.0 drop. Here's what broke, why, and which version to pin.
MCP Credential Leak Closed in Claude Code's Busiest Week Yet
Seven builds in one week: four Bash/PowerShell sandbox bugs patched, /code-review --fix lands auto-apply, and a serious MCP auth credential leak is closed.
Anthropic 0.105.0 Adds Output Attribution — What It Buys You
v0.105.0 adds granular output-type attribution and configurable upload caps—here's what they do and when to use them.
The Claude Code Sprint That Patched Four Security Holes
Ten patches in nine days: pinned sessions, four security fixes, /code-review --fix, and skill-level tool gating.
Anthropic Python SDK 0.105: Opus 4.8 and Mid-Session System Prompts
Three SDK releases in 7.5 hours ship claude-opus-4-8 support, mid-conversation system blocks, and finer output usage reporting.
Claude Opus 4.8 Hits 69.2% SWE-Bench Pro — What Else Changed
Anthropic ships Opus 4.8 with 69.2% SWE-Bench Pro, mid-conversation system messages, and adaptive thinking.
Anthropic Pays a Rival $1.25B/Month. Claude Rate Limits Jump.
Anthropic buys exclusive access to xAI's Colossus 1 cluster: 220K GPUs, $1.25B/month, and immediate Claude rate limit increases.
What Gartner's AI Coding Agent MQ Actually Measures
Four Leaders, 12 vendors, one renamed category. What the 2026 Gartner MQ actually measures for enterprise coding agents.
260억 달러 유치 후 AI 코딩 에이전트 지형이 달라졌나
GitHub Agent HQ부터 Devin의 260억 달러 투자 유치까지 — 2026년 AI 코딩 에이전트의 변화를 기술적으로 해부한다.
대화 중간 제약을 바꿔도 프롬프트 캐시가 끊기지 않는다
Mid-conversation constraint injection in v0.105.0 preserves prompt cache continuity across long inference runs.
Anthropic SDK 릴리즈가 PyPI 배포를 깨뜨린 이유
Two rapid patches followed Anthropic's 0.105.0 drop. Here's what broke, why, and which version to pin.
Claude Code MCP 크리덴셜 유출이 패치됐다
Seven builds in one week: four Bash/PowerShell sandbox bugs patched, /code-review --fix lands auto-apply, and a serious MCP auth credential leak is closed.
Anthropic SDK 출력 귀속, 코드에서 실제로 뭐가 달라지나
v0.105.0 adds granular output-type attribution and configurable upload caps—here's what they do and when to use them.
Claude Code, 9일 만에 보안 구멍 4개를 닫았다
Ten patches in nine days: pinned sessions, four security fixes, /code-review --fix, and skill-level tool gating.
Anthropic Python SDK 0.105: Opus 4.8 및 미드-세션 시스템 프롬프트
Three SDK releases in 7.5 hours ship claude-opus-4-8 support, mid-conversation system blocks, and finer output usage reporting.
SWE-Bench Pro 69.2%의 Claude, 에이전트 코딩이 달라지나
Anthropic ships Opus 4.8 with 69.2% SWE-Bench Pro, mid-conversation system messages, and adaptive thinking.
xAI Colossus 독점 계약으로 Claude 요청 한도가 즉시 올랐다
Anthropic buys exclusive access to xAI's Colossus 1 cluster: 220K GPUs, $1.25B/month, and immediate Claude rate limit increases.
AI developer tools and ecosystem news for developers and technical founders
