Daily briefs covering Claude Code releases, Anthropic drops, MCP ecosystem changes, and the broader AI news that actually changes how builders ship. Published every morning.
Claude Code 2.1.160 to 2.1.162 ship config-write guardrails and an MCP timeout fix; Anthropic maps a year of AI-enabled cyber threats (33% to 56% actor adoption) and expands Project Glasswing with Claude Security.
Claude Code 2.1.160/161 add write-guard prompts and fix an MCP secret leak; Anthropic expands Project Glasswing to ~150 orgs and publishes a year-in-review on AI-enabled cyber threats; MCP's sessionless 2026-07-28 spec RC stays in flight.
Claude Code 2.1.160 guards acceptEdits config writes and renames the dynamic-workflow keyword to ultracode, Auto mode reaches Bedrock/Vertex/Foundry, Anthropic files its draft S-1, and the June 15 programmatic-credit split nears.
Claude Code 2.1.158/159 ship Auto mode on Bedrock, Vertex, and Foundry for Opus 4.7/4.8; researchers bypass every tested AI defense above 90 percent; MCP stateless-core RC locks ahead of the July 28 spec.
Claude Code 2.1.158 extends auto mode to Bedrock/Vertex/Foundry; 2.1.157 auto-loads .claude/skills plugins; Opus 4.8 ships high-effort default plus 2.5x fast mode; MCP 2026-07-28 spec RC locks a stateless core.
Claude Code 2.1.158 brings auto mode to Bedrock/Vertex/Foundry; Opus 4.8 lands at 88.6% SWE-bench with Dynamic Workflows; MCP Python/Go/TypeScript SDKs all refreshed for the Tasks extension.
Claude Opus 4.8 ships as the default in Claude Code 2.1.154 with Dynamic Workflows (capped at 1,000 subagents), a cheaper fast mode, /code-review --fix, and Anthropic's $65B raise at a $965B valuation.
Claude Code 2.1.153 ships skipLfs and status-line env vars, 2.1.152 adds MessageDisplay hook and disallowed-tools frontmatter, BadHost CVE-2026-48710 hits Starlette-based MCP gateways, Malware-Slop npm package targets Claude's /mnt/user-data.
Claude Code 2.1.152 ships /code-review --fix and MessageDisplay hook, Anthropic launches Compliance API with 28 enterprise security integrations, AWS MCP Server hits GA.
Anthropic ships 28 security/compliance integrations for Claude, Mythos surfaces in Claude Code, Opus 4.7 hits elevated errors, MCP 2026-07-28 RC governance set.
Claude Code 2.1.150 ships, Project Glasswing reports 10,000+ vulns via Claude Mythos Preview, Claude Security hits public beta, and the MCP 2026-07-28 release candidate lands with a stateless core.
Claude Code 2.1.149 ships per-MCP /usage breakdown and PowerShell sandbox fixes; Anthropic posts Project Glasswing update with Claude Security beta; MCP 2026-07-28 spec locked as release candidate.
Claude Code 2.1.149 ships /usage breakdown and a PowerShell sandbox escape fix, /simplify becomes /code-review with PR comment posting, MCP 2026-07-28 release candidate locks with a stateless core, and Anthropic posts a Project Glasswing update.
Claude Code 2.1.148 hotfixes bash exit-code regression, 2.1.147 ships pinned background sessions and a rebuilt /code-review command, Anthropic widens frontier-AI conversation, MCP ecosystem absorbs AWS GA server lessons.
Claude Code 2.1.146 ships /code-review rename plus MCP pagination fix, Anthropic opens moral-formation dialogue and KPMG rollout, MCP stateless HTTP transport advances toward June spec.
Claude Code 2.1.145 ships scriptable agent sessions and OTEL agent_id spans, Anthropic acquires Stainless to own SDK and MCP server generation, and an ethical-reminder tool cuts misalignment in eval runs.
Anthropic acquires Stainless for SDK and MCP tooling, Claude Code 2.1.144 ships background /resume and pagination fix, 20+ legal MCP connectors launch.
Quiet weekend for shipped Claude Code and Anthropic releases. MCP TypeScript SDK saw repo activity on May 17. PraisonAI CVE-2026-44338 keeps demonstrating the new exploit clock for agent frameworks.
Claude Code 2.1.143 ships plugin guards and projected context cost, Anthropic publishes 2028 AI leadership scenarios, programmatic Claude usage moves to its own credit meter June 15.
Claude Code 2.1.143 lands plugin dependency enforcement and projected context cost; fast mode defaults to Opus 4.7; Anthropic announces split billing for Agent SDK effective June 15; Claude Platform GA on AWS; MCP 2026 roadmap targets Server Cards and stateless transport.
Claude Code 2.1.142 ships agents flag suite and Opus 4.7 fast mode, Anthropic splits programmatic billing June 15, Gates Foundation $200M partnership, 12 new legal practice tools.
Claude Code 2.1.141 ships workload identity federation and Rewind compression, Anthropic launches Claude for Small Business plus 20+ legal MCP connectors, Claude Code weekly limits bump 50% through July 13.
Claude Platform launches on AWS with native API access, Claude for Legal ships 20+ MCP connectors and 12 practice plugins, Claude Code 2.1.140 fixes /loop and subagent matching.
Claude Code 2.1.139 ships Agent view and /goal autoloop, Google confirms first AI-built zero-day exploited in the wild, Microsoft patches RCE in Semantic Kernel
Quiet weekend window. Claude Code v2.1.136 to v2.1.138 fix MCP session persistence after /clear, OAuth refresh races, and add settings.autoMode.hard_deny.
Claude Code 2.1.136-138 fixes MCP config persistence and OAuth token loss; Anthropic reports 80x quarterly growth and $1.8B Akamai compute deal; Claude misused in Mexican water utility intrusion attempt.
Claude Code 2.1.136-138 ships hard_deny auto mode and 50+ MCP/OAuth fixes; Anthropic publishes 'Teaching Claude Why' on principle-based alignment; Dragos discloses Claude-driven OT intrusion at Mexican water utility.
Code w/ Claude 2026 ships Routines, Multi-Agent Orchestration, and Dreaming preview; Claude Code 2.1.133 lands worktree.baseRef and effort-level hooks; AWS MCP Server hits GA; Anthropic doubles Pro/Max rate limits and signs SpaceX Colossus 1 deal.
Claude Code 2.1.132 ships, Routines and Remote Agents launch, Claude Agent SDK opens externally, Anthropic plus SpaceX 300MW deal doubles Pro/Max rate limits, AWS MCP Server hits GA.
Claude Code 2.1.131 ships Windows VS Code fix; Anthropic launches 10 finance agent templates with Moody's MCP app and Microsoft 365 add-ins; Memory on Managed Agents enters public beta.
Claude Code v2.1.128 ships plugin zip support and quieter MCP reconnects; Anthropic launches enterprise services JV with Blackstone, H&F, Goldman; Jama releases first engineering-management MCP server.
Claude Security public beta hits broader coverage on Opus 4.7, Claude Code 2.1.126 ships gateway model picker and MCP auto-retry, MCP 2026 roadmap locks four priorities, Comment and Control disclosure earns CVSS 9.4 across coding agents.
Claude Code 2.1.126 ships project purge and gateway model picker, Anthropic retires 1M context beta on Sonnet 4.5/4, MCP SDKs (Python, Go, Ruby) and Inspector all cut releases.
Claude Code 2.1.126 ships project purge and gateway model picker; Claude Security hits public beta on Opus 4.7; MCP variants reference impl lands across SDKs.
Anthropic ships Claude Security public beta on Opus 4.7, Claude Code 2.1.126 lands gateway model picker and project purge, MCP roadmap centers on Streamable HTTP and Server Cards.
1M context beta retires today on Sonnet 4 and 4.5; Anthropic joins Blender Development Fund as Corporate Patron at €240k/year; Blender Lab ships MCP server.
Claude Code 2.1.121-123 ships alwaysLoad MCP config, Auto mode for Max, xhigh effort, /ultrareview; Anthropic launches Claude for Creative Work with 9 connectors; Affinity ships hosted MCP server.
Claude Code v2.1.120 and v2.1.121 ship same-day with Windows PowerShell fallback, MCP alwaysLoad, claude ultrareview CLI, and multi-GB memory leak fixes; AWS and Anthropic deepen silicon co-engineering with Claude Cowork live in Bedrock.
Claude Code v2.1.119 ships vim visual mode, /config persistence, and MCP OAuth fixes; 30+ regressions filed against v2.1.119/120 on April 24; Google commits up to $40B to Anthropic.
Claude Code 2.1.120 ships BrowserBatch, persistent memory, and headless ultrareview, but a --resume crash and seven other regressions have auto-update reverting clients to 2.1.119; Anthropic and NEC partner to roll Claude to 30,000 employees; MCP spec unchanged.
Anthropic's April 23 postmortem traces Claude Code's monthlong quality slump to three specific changes (reasoning effort cut, idle-session memory bug, 25-word response cap), all fixed by v2.1.116; election safeguards update lands with Opus 4.7 / Sonnet 4.6 neutrality scores; no new Claude Code release in 48 hours.
Claude Code 2.1.119 ships 40+ fixes with MCP OAuth repairs and agent frontmatter honored in --print; Claude Managed Agents adds built-in memory in public beta; Anthropic and NEC ink Japan's largest AI-native engineering rollout across 30,000 staff.
Claude Code 2.1.118 adds vim visual mode, MCP tool hooks, and named themes; 2.1.117 lands forked subagents and model persistence; Anthropic walks back quiet Claude Code pricing A/B test.
Claude Code 2.1.117 ships with agent-frontmatter MCP loading, forked subagents, and high-effort default on Pro/Max. Bloomberg reports an unauthorized group accessed Claude Mythos through a third-party vendor. Anthropic quietly tests pulling Claude Code from the Pro plan.
Claude Code 2.1.116 cuts /resume time by 67% and closes a sandbox rm hole, Amazon commits up to $25B more to Anthropic with 5GW of compute runway, and the MCP spec lands OAuth 2.1 with incremental scope consent.
Claude Haiku 3 (claude-3-haiku-20240307) retires today. No new Claude Code release, no Anthropic news, MCP upstream still quiet. If your pipeline still calls Haiku 3, it's already broken.
Quiet 48 hour window: Claude Code 2.1.114 ships a single Agent Teams permission-dialog crash fix, no new Anthropic releases, and MCP upstream stays silent.
Claude Code 2.1.113 ships a native binary and tighter Bash sandboxing, Opus 4.7 lands with xhigh effort and task budgets, and Anthropic launches Claude Design.
Launching a daily AI briefing for builders: Claude Code changelog, Anthropic drops, MCP ecosystem, and the short list of broader AI news that actually changes how you ship.