Claude Report — 2026-06-30
- Anthropic launches self-hosted Claude apps gateway for Bedrock and Google Cloud — corporate SSO, RBAC, per-user cost attribution, and spend limits for Claude Code teams (Jun 29)
- Claude in Microsoft Foundry is now GA: Opus 4.8 + Haiku 4.5 on Azure, running on NVIDIA GB300 Blackwell Ultra GPUs (Jun 29)
- Claude Code v2.1.196: org default models, ~25% cheaper `/code-review`, stream idle watchdog on by default, security fix stopping `.mcp.json` server auto-spawn (Jun 29)
- `claude-opus-4-6` fast mode silently retired — requests now run at standard speed and standard billing without error (Jun 29)
- anthropic-sdk-python v0.113.0 adds `20260318` web-fetch and support tools; claude-agent-sdk-typescript v0.3.196 adds OTel hook correlation (Jun 29)
🚀 Anthropic Official
Claude apps gateway for Amazon Bedrock and Google Cloud (Jun 29)
Ships a self-hosted, single-container control plane for Claude Code on Bedrock and Vertex: adds corporate SSO via OIDC (Google Workspace, Microsoft Entra ID, Okta), centrally pushed managed settings, per-user cost attribution, and spend limits at org/group/user levels with optional failover routing between providers.
Claude in Microsoft Foundry is now generally available (Jun 29)
Claude Opus 4.8 and Claude Haiku 4.5 land in Microsoft Foundry on Azure, running on NVIDIA GB300 Blackwell Ultra GPUs; two deployment modes offered — Azure-hosted (data residency, existing Azure billing) and Anthropic-hosted (full feature set) — live in East US and West Europe today, Southeast Asia/Australia East/UK South by mid-July.
API: claude-opus-4-6 fast mode removed (Jun 29)
Requests to claude-opus-4-6 with speed: "fast" now run at standard speed and are billed at standard rates with no error returned; usage.speed reports the actual speed used; migrate to Claude Opus 4.8 to keep fast mode.
Claude Code v2.1.196 (Jun 29)
Adds org default models (admins set in the console; shows as "Org default"/"Role default" in /model), readable auto-generated session names, and Cmd/Ctrl-click to reveal attached files in Finder/Explorer; reduces /code-review token usage by ~25% and enables the stream idle watchdog by default (aborts/retries after 5 minutes with no events); security: claude mcp list/get no longer spawn servers from repo-committed .mcp.json, and untrusted workspaces show "⏸ Pending approval"; fixes background job data loss, PowerShell git diff/git grep failures, MCP OAuth scope handling on GitLab self-hosted, and multiple claude agents panel issues.
anthropic-sdk-python v0.113.0 (Jun 29)
Adds support for 20260318 web-fetch and support tools, accepts user profile IDs in count_tokens, and fixes a missing output_format/output_config merge block in async count_tokens.
🔌 Claude Code Plugins
Plugin tracking skipped — ranking cache unavailable this cycle.
🛠️ Skills
None of the top 5 updated in window.
Unchanged in window: anthropics/skills, alirezarezvani/claude-skills, Piebald-AI/claude-code-system-prompts, glebis/claude-skills, ComposioHQ/awesome-claude-skills
🤖 Agents & Subagents
claude-agent-sdk-typescript v0.3.196 (Jun 29)
Adds prompt_id field to hook input payloads for correlating hook events with OpenTelemetry prompt-level spans; fixes control-protocol deduplication dropping tool-use IDs after 1,000 resolutions, which could cause duplicate tool_result deliveries in long-running sessions.
Unchanged in window: nicobailon/pi-subagents, Chachamaru127/claude-code-harness
🔗 MCPs & Integrations
None of the top 5 updated in window.
Unchanged in window: github-mcp-server, mcp-python-sdk, awslabs/mcp, modelcontextprotocol/typescript-sdk, portainer/portainer-mcp
💡 Community — Workflows & Ideas
shanraisshan/claude-code-best-practice updated (Jun 30)
The 61.6k-star Claude Code reference guide received its June 30 auto-commit via Claude Code itself — a practical illustration of the tool updating its own documentation — incorporating v2.1.196 changes including org default model behavior, stream idle watchdog defaults, and /code-review efficiency patterns across its 83-tip catalog and command→agent→skill orchestration examples.
📰 Quick Mentions
Relevant links that don't warrant a section of their own:
- NVIDIA: Anthropic's Models Now Run on GB300 Blackwell Ultra in Azure (Jun 29) — NVIDIA's perspective on the Foundry GA: first Anthropic deployment on NVIDIA hardware, performance benchmarks for Opus 4.8
- Azure Blog: Claude in Microsoft Foundry GA (Jun 29) — Microsoft's announcement with enterprise architecture details, US data zone info, and customer case studies from NVIDIA, Bolt, Everstar, and Momentic