Summary
The Week 25 (June 2026) release of Claude Code (versions v2.1.178‑183) introduces Artifacts — live, interactive pages published from a session to a private URL on claude.ai that update in place as the session works. It also adds permission rules with input‑parameter matching, the ability to set any config key from the prompt via /config key=value, deeper subagent nesting (up to 5 levels), and --safe-mode for disabling customizations.
Key Points
- Artifacts: Shareable, live-updating pages published from a session; replaces using the terminal for things like PR walkthroughs or dashboards.
- Permission rules now match tool input parameters using
Tool(param:value)syntax (e.g.,Agent(model:opus),Agent(isolation:*)). /config key=valuesets any configuration setting from the prompt, also works non‑interactively with-p.- Subagents can spawn subagents (up to 5 levels deep).
--safe-modedisablesCLAUDE.md, skills, plugins, hooks, MCP, and custom commands.
Concepts
- Artifact: A live page (HTML/JavaScript/CSS/react?) published from a Claude Code session to a private URL on claude.ai; it updates in real‑time as the session continues. Currently in beta for Team and Enterprise plans.
- Permission Rule Parameter Matching: Extends tool‑name‑only permission rules to also filter by the input parameters supplied to a tool, using a
Tool(param:value)pattern. - Subagent Chain: A hierarchy of spawned subagents, now allowed up to 5 levels deep (previously fewer).