Source: raw/anthropic-watch-claude-code-tag-v2-1-284.md (Claude Code v2.1.284, published 2026-09-28) and raw/anthropic-watch-skills-commit-8a1541c4a3ffa5a20a5a91de0dcf3f0bab1d1ef4.md (anthropics/skills, 2026-09-29 02:20 UTC), with the same-day SDK releases raw/anthropic-watch-anthropic-sdk-python-tag-v1-9-0.md and raw/anthropic-watch-anthropic-sdk-typescript-tag-sdk-v0-129-0.md. Additional releases v2.1.285–v2.1.288 compiled from ai-research/claude-code-docs-changelog-2026-10-02.md.

Product: Claude Code · Date: 2026-09-28 to 2026-10-02

Week 40 covers six releases (v2.1.284–v2.1.289, September 28–October 3, 2026). The week opened with Claude Sonnet 5.5 as the new default Sonnet on the Anthropic API and auto mode as the universal default permission mode. It continued with new CLI tools (claude --desktop, claude plugin configure, claude purge), a new admin control (allowedProviders), the launch of Claude Mods — a new plugin capability type that lets plugins modify Claude Code’s own internal behavior, shipping its first surface ($.ui.selection()) and a built-in Mod (“You should know”) — and closed with agent.spawn for teammates (v2.1.289, October 3). Updated 2026-10-09 to add v2.1.289. Continues Week 39.

Key Takeaways

  • Claude Sonnet 5.5 (claude-sonnet-5-5) is the default Sonnet model on the Anthropic API — 1M context, 10 per Mtok, $0.20/Mtok cache reads (v2.1.284, 2026-09-28). See Claude Sonnet 5.5.
  • Auto mode is now the starting permission mode everywhere. Interactive terminal and VS Code sessions start in auto mode when no permission mode is configured, on every plan and provider; permissions.defaultMode still overrides it (v2.1.284). See Auto Mode.
  • Ultracode is its own toggle in /effort (Tab, or /effort ultracode [on|off]): it no longer forces xhigh effort and stays on at any effort level (v2.1.284). See Dynamic Workflows.
  • {"decision":"block"} from Elicitation and ElicitationResult hooks now declines the MCP elicitation, as exit code 2 does; it was being ignored (v2.1.284). See Hooks.
  • The claude-api skill makes Opus 5.5 the default model, adds Sonnet 5.5, and ships build-eval and hillclimb flows (skills 8a1541c4).
  • SDKs added claude-sonnet-5-5 the same day — Python v1.9.0, TypeScript sdk-v0.129.0. See Anthropic SDK Releases Aug–Sep 2026.
  • claude --desktop opens the Claude Desktop app on the current directory from the CLI (v2.1.285). See CLI Reference.
  • allowedProviders managed setting limits which API providers a machine may use (v2.1.285).
  • CLAUDE_CODE_DISABLE_WEB_FETCH env var turns off the WebFetch tool entirely (v2.1.285).
  • claude plugin configure <plugin> shows a plugin’s options and which are unset; --values-stdin saves new values (v2.1.285).
  • Background Bash/PowerShell commands now stop after a time limit (default 30 min, max 2 h) in unattended sessions; terminal/desktop/VS Code sessions have no limit — changed in v2.1.285/v2.1.288.
  • Claude Mods launched in v2.1.287: a new plugin capability type letting plugins modify Claude Code’s own behavior loop. Built-in mod “You should know” watches your session and surfaces overlooked items. Enable: /plugin enable cc-plugin-you-should-know@builtin.
  • $.ui.selection() (v2.1.288): the first Mods API call, returning the user’s last selected text in fullscreen mode.
  • --max-findings <n>|all added to /code-review (v2.1.288). See CLI Reference.
  • claude purge replaces claude project purge; old form still works and prints a notice (v2.1.288).
  • Auto mode compacts long conversations instead of prompting or failing every tool call when the conversation is too long for the client-side safety classifier (v2.1.288).

Release table

VersionDate (UTC)Shape
v2.1.2842026-09-28Claude Sonnet 5.5; auto mode default for all; Ultracode toggle; /mcp reconnect all
v2.1.2852026-09-29claude --desktop; allowedProviders; CLAUDE_CODE_DISABLE_WEB_FETCH; claude plugin configure; background Bash time limit; 70+ fixes
v2.1.2862026-09-30Permission prompt stacking count (“N of M”); mouse support for overflow rows; 40+ fixes; VSCode bookmarks
v2.1.2872026-10-01Claude Mods launch; “You should know” built-in mod; URL prompts from MCP; agents view n: filter; prompt_text OTel field
v2.1.2882026-10-02$.ui.selection() for mods; --max-findings for /code-review; claude purge; Ctrl+F session search; auto mode compacts on classifier overflow; 50+ fixes
v2.1.2892026-10-03agent.spawn for teammates; idle/waiting states in $.agent.list(); VS Code auth revert; 23 fixes

Models and defaults

  • Claude Sonnet 5.5 added, default Sonnet on the Anthropic API (v2.1.284).
  • Auto mode default on every plan and provider when no permission mode is set (v2.1.284).
  • Ultracode decoupled from effort level (v2.1.284).
  • Safety-related model switch target chosen by the API per kind of flag when an Opus pin is set (v2.1.284).
  • The notice shown when a Sonnet model’s safeguards flag a message now explains why and offers editing and retrying (v2.1.284).
  • The Explore subagent no longer switches to Opus on the Claude API when the session runs an unrecognized model ID, such as a custom model behind a proxy; Explore inherits that model (v2.1.284). See Subagents.

New commands and features

  • /mcp reconnect all retries every MCP server that failed to connect or needs authentication (v2.1.284). See MCP.
  • /rate-limit-options appears in /help and the command menu for claude.ai subscribers (v2.1.284).
  • Auto mode’s prompt before a read outside the working directories gains “Yes, but ask again next time” (v2.1.284).
  • Gateway spend limits show dollar amounts in /usage and the status line (e.g. “500.00 spent this month”); rate_limits.spend_limit gains used_usd, limit_usd and period (v2.1.284). See Claude Apps Gateway.
  • Gateway: startup warnings when a managed policy’s availableModels is empty or omits the starting model; auth: { google: {} } for telemetry export to Google Cloud’s OTLP endpoint; certificate client authentication (private_key_jwt) with the identity provider (v2.1.284).
  • /claude-api hillclimb no longer spends rounds on prompt rewordings too small for the eval to measure (v2.1.284).
  • Monitor event rows show what each event printed instead of repeating the description (v2.1.284). See Monitor Tool.
  • Claude Tag: model family choices such as “Opus (latest)” for a thread, a channel default or a DM, so the choice follows the newest model in that family (v2.1.284). See Claude Tag.

Hooks, MCP, plugins and skills

  • Elicitation / ElicitationResult {"decision":"block"} honoured (v2.1.284).
  • A failed hook’s stderr is no longer dropped from the debug log when it also wrote stdout, and failed hooks log their status code (v2.1.284).
  • MCP tool calls in a resumed session wait up to 10 seconds for a still-connecting server instead of failing with “No such tool available” (v2.1.284).
  • claude mcp add refuses when managed settings restrict MCP servers to plugins (v2.1.284).
  • Plugins from marketplaces, claude.ai and npm no longer pre-approve their own tools via allowed-tools under managed allowManagedPermissionRulesOnly; only official Anthropic sources, or sources managed settings vouch for, keep that pre-approval (v2.1.284). See Plugins and Marketplaces.
  • The non-interactive first turn still waits up to 2 seconds for MCP servers named by --allowedTools or an mcp_tool hook, even when CLAUDE_CODE_MCP_STARTUP_WAIT_MS is 0 (v2.1.284).
  • /loop self-paced mode writes each status update, and the outcome when it stops, as visible text instead of only in reasoning (v2.1.284). See Scheduled Tasks.
  • Rules symlinked into .claude/rules from outside the project now trigger the external-imports approval prompt (v2.1.284).
  • Auto-memory: invisible characters and tags imitating Claude Code’s own markup are neutralized in MEMORY.md and recalled notes (v2.1.284). See Auto Memory.

The claude-api skill update (skills 8a1541c4, 2026-09-29)

  • Claude Opus 5.5 (claude-opus-5-5) becomes the default model across SKILL.md, models.md and every language’s examples, with Opus 5 as the previous Opus; a “Migrating to Claude Opus 5.5” section is added to shared/model-migration.md.
  • Claude Sonnet 5.5 (claude-sonnet-5-5) is the current Sonnet, with Sonnet 5 as the previous one; pricing, model-support callouts and platform availability updated.
  • Build-eval and hillclimb flows under shared/evals/ (build-eval, eval-hillclimb, cost-hillclimb, eval-audit), with a report schema, a lite HTML report builder and a runner scaffold.
  • A preserved-thinking-migration subcommand with probe scripts, and a note that drop_block can raise token usage; cost-optimize updated for preserved thinking and per-model cache rates, pointing cache misses at cache diagnostics.
  • Streaming client tools default to eager_input_streaming; model access failures documented as a 404 and beta gating as a 400, not a 403.
  • ant apply recommended for version-controlled Managed Agents resources. See Managed Agents.

v2.1.285 — New CLI and admin additions (September 29)

  • claude --desktop opens the Claude Desktop app on the current directory; with --continue or --resume <id> it opens on a session.
  • claude plugin configure <plugin> shows a plugin’s options and which are unset; --values-stdin saves new values from stdin. <server>.<key>=<value> on claude plugin install --config lets bundled .mcpb server settings be set at install time.
  • allowedProviders managed setting limits which API providers a machine may use: Anthropic API, a custom endpoint, Bedrock, Mantle, Vertex AI, Foundry, Claude Platform on AWS, or a Cloud gateway.
  • CLAUDE_CODE_DISABLE_WEB_FETCH turns off the WebFetch tool for the session.
  • claude mcp list now includes WebSocket (ws) MCP servers, each shown with its URL and health status.
  • Sessions behind a custom ANTHROPIC_BASE_URL now use the 1M context window for models that have one (Opus 4.7+, Sonnet 5+, Fable).
  • Background Bash/PowerShell commands stop after a time limit (default 30 min, max 2 h) in unattended sessions; claude -p, Agent SDK, CI and cloud sessions have the limit. Terminal, desktop and VS Code sessions are unlimited (changed further in v2.1.288).
  • /claude-api can no longer be run from Remote Control clients.
  • Fixes: claude mcp get hides command and args for plugin-provided stdio servers; URL passwords and credentials are better redacted in logs; resumed sessions no longer lose compacted context.

v2.1.286 — Permission prompt stacking and fixes (September 30)

  • “N of M” count on stacked permission prompts: when several permission requests are queued, the prompt now reads “2 of 5” so you know how many are waiting.
  • Mouse support for “N more” rows in fullscreen lists: click to jump to that end of the list.
  • Waiting permission prompts now show oldest first, so a new prompt no longer covers the one you’re reading.
  • Fixes: multiple crashes in --resume and --continue on sessions with batched parallel tool calls; API 400 errors when a hook returned a non-string value; cloud sessions with very large histories failing to wake; spend meter mispricing 1-hour cache writes; macOS sessions staying “Not logged in” after a successful /login in another window.
  • Fixes: /compact, /clear, and /rewind in a subagent’s transcript now ask which target to act on instead of silently acting on the main conversation.
  • VSCode: bookmarks (save Claude’s responses to a Bookmarks side panel), question cards with option previews, rows under a message showing terminal output and browser state.

v2.1.287 — Claude Mods launch (October 1)

Claude Mods is the headline feature of v2.1.287. Mods are a new plugin capability type that let a plugin modify Claude Code’s own internal behavior loop, not just add tools or panels.

  • “You should know” (cc-plugin-you-should-know@builtin): Anthropic’s built-in Mod. A side agent watches your session and surfaces things you or Claude might miss. Enable: /plugin enable cc-plugin-you-should-know@builtin. Requires a first-party session (Anthropic API login) with telemetry on.
  • URL prompts from MCP servers on the 2025-11-25 protocol (for sign-in flows). If a server no longer connects after this update, add "bareElicitationCapability": true to its MCP config entry.
  • n:<text> filter in the agents view matches session names and tasks; Enter opens the first match.
  • prompt_text added to the OpenTelemetry user_prompt event (copy of prompt for backends that nest dotted keys; mask or drop wherever you drop prompt).
  • Opus 4.7+ and Fable now use the 1M context window by default on Bedrock, Vertex, Foundry and the Claude apps gateway (no [1m] suffix required; CLAUDE_CODE_DISABLE_1M_CONTEXT=1 keeps 200K).
  • Fixes: Sonnet 5.5 can now advise Opus 4.7 and 4.8 in /advisor; Bedrock/Vertex startup checks no longer ignore an enforced availableModels list; hooks configured with asyncRewake no longer wake Claude repeatedly when the hook script is missing.

v2.1.288 — Mods API surface and major fixes (October 2)

  • $.ui.selection(): the first Mods API call. Returns the text last selected in fullscreen mode; when the selection falls within one transcript row, returns that row’s data.
  • --max-findings <n>|all for /code-review: report more or fewer findings than the default limit. The choice is reused until you pass --max-findings default.
  • claude purge replaces claude project purge; the old form still works and prints a migration notice.
  • Ctrl+F in the agents view finds a session by name; Alt+↑/↓ jumps between groups.
  • Auto mode compacts a conversation that has grown too long for the client-side safety classifier, instead of prompting or failing every tool call.
  • Built-in gh api added to cloud sessions whose container image has no GitHub CLI installed.
  • CLAUDE_CODE_DISABLE_STRUCTURED_OUTPUTS turns off structured outputs for session titles, memory recall and prompt hooks — a workaround for Mantle and certain gateway backends.
  • Fixes: mid-response API timeouts in non-interactive sessions and subagents now continue from the partial response; --resume no longer drops compacted context; plugin-defined agents now run with their own prompt, tools and effort; git-subdir plugin installs work on Ubuntu 22.04 (git 2.34); a dangerous rm on / or ~ inside bash -c is now stopped in bypassPermissions mode; LSP tool calls time out after 60s instead of hanging.

Breaking and behaviour changes

  • /recap declines with a short notice when relayed from a chat thread, routine or webhook (v2.1.284).
  • Retries after a dropped connection mid-response share one budget with the request’s other retries, so a failing request gives up sooner (v2.1.284).
  • Artifact publishing refuses a file on a network share unless it is on a mapped drive added with --add-dir (v2.1.284).
  • VS Code: CLAUDE_CONFIG_DIR in claudeCode.environmentVariables applies only when it is an absolute path (v2.1.284).

Notable fixes

  • “Prompt is too long” persisting after compaction: Claude Code now compacts once more, keeping less of the recent conversation (v2.1.284).
  • An overloaded or server error right after a thinking block is retried instead of ending the turn (v2.1.284).
  • ANTHROPIC_FOUNDRY_RESOURCE was interpolated into the Foundry endpoint host unvalidated; non-plain values are now refused (v2.1.284).
  • Claude Desktop behind a Claude apps gateway offered no 1M context option; the gateway now marks each 1M-capable model (v2.1.284).

Try It

  • If you don’t want to start in auto mode, set permissions.defaultMode in user or managed settings.
  • Pinned Opus via ANTHROPIC_DEFAULT_OPUS_MODEL? A safety-related switch may now land on a model you didn’t pin; check the notice when it happens.
  • Using Ultracode at lower effort: turn it on with /effort ultracode on and set effort separately.
  • Update the bundled skill’s examples in your own code to claude-opus-5-5 / claude-sonnet-5-5 if you copied model IDs from the claude-api skill.

v2.1.289 — Agent spawn and teammate states (October 3)

Source: ai-research/claude-code-docs-changelog-2026-10-09.md.

  • agent.spawn for teammates: spawn new agent teammates dynamically in a running session rather than only at session initialization.
  • Idle and waiting states in $.agent.list(): the agent list now returns idle and waiting as distinct states, giving finer-grained visibility into teammate lifecycle.
  • Reverted: a claude auth status change introduced in v2.1.288 that was breaking VS Code auth display.
  • 23 bug fixes.

See Claude Code Agent Teams for teammate management patterns.

Open Questions

  • Is Sonnet 5.5 the default Sonnet on Bedrock, Vertex and Foundry? v2.1.284 says “on the Anthropic API” only. The TypeScript Bedrock, Vertex and Foundry SDKs added the model ID the same day; the change to Bedrock/Vertex/Foundry defaults was not stated in any v2.1.285–289 note.
  • What does the Opus 5.5 migration section say? The skills commit names the section but does not reproduce its contents.
  • What Mods API surfaces follow $.ui.selection()? The first Mods API call shipped in v2.1.288; no registry of planned surfaces is published.
  • Do third-party Mods require telemetry on? The “You should know” built-in requires it; whether this applies to all Mods is not stated.
  • Resolved 2026-10-09: v2.1.289 (October 3, 2026) was published within the W40 calendar window — agent.spawn for teammates and idle/waiting states. See section above.