1.0.74-4

Added

• Add support for Open Plugin Spec v1 plugin manifests and mcp.json configuration

Improved

• Subagent timelines identify whether prompts came from the main agent or another subagent

Fixed

• IDE integration reconnects reliably when the CLI reloads MCP servers or changes directory
• Multi-turn subagent timelines show every prompt and response in the correct order after reopening /tasks

1.0.74-3

Fixes and changes

1.0.74-2

Fixes and changes

1.0.74-1

Added

• Show a first-run splash to opt into the default sandbox
• Adding support for gemini-3.6-flash

Improved

• When multiplexing sessions, a session's open dialog no longer leaks into another session; eligible pickers reopen when you switch back
• The $ interactive shell shortcut now opens a shell even while the agent is working
• Fully honor the skill disable-model-invocation flag
• Warn when a participating language server reports a different symbol than the one requested
• Steering interrupts shell output waits without stopping the running command
• Increase the Responses request size limit
• Plan mode now allows session-folder planning artifacts while still blocking clear file mutations outside the session folder.

Fixed

• The /mcp add and /mcp edit wizard now preserves = characters in environment variable values (such as base64 padding), so secrets and tokens are stored correctly.
• Remote session uploads stop retrying permanent Mission Control 400/404 responses
• Show Tab in /settings footer to switch scope tabs
• Downscale oversized tool-result images so CAPI Responses requests continue

1.0.74-0

Added

• Add /model plan (or /model --plan) to pick a model used while in plan mode; pass a model id, off to clear, or no id to open the picker. Reverts to the session model when you leave plan mode.

Improved

• Resume search matches session titles even when whitespace differs

1.0.73

2026-07-20

• Anthropic subagents continue working when additional directories are configured
• Resolve relative links in custom agent instructions from the agent file location

1.0.72

2026-07-20

• An agentStop hook that always blocks no longer loops indefinitely: the CLI now ends the turn after 8 consecutive blocks, and agentStop hooks receive a stop_hook_active flag so they can detect a forced continuation and self-limit
• Add opt-in git and gh authentication inside the OS sandbox
• Sandbox macOS keychain access now defaults off for tighter isolation; re-enable it in /sandbox if a command needs it
• Lifecycle and subagent hook commands run in the current session directory after /cd
• Deleting an MCP server with /mcp delete stops its running background process
• Toggling /sandbox restarts only local MCP servers and leaves remote servers connected
• Command approvals no longer carry over to another repository after you switch with /cd
• The GitHub tab's Open in web action now reliably launches your browser on Windows
• Preserve pasted prompt content when changing models through Ctrl+X /model
• Corrected the added-line count shown when creating a file so files ending in a trailing newline no longer report one extra line
• /worktree and /move no longer fail to create a worktree for an auto-generated branch name when many similarly-named branches or a leftover worktree directory already exist: the numeric-suffix search is no longer capped at 5 and now skips stale unregistered worktree directories (an explicitly supplied branch name still errors on a collision)
• /worktree no longer intermittently runs the kickoff task in the main repo instead of the new worktree
• /worktree and /move now propagate folder trust to the new worktree before switching when the source is already trusted (avoiding a spurious folder-trust prompt in that case), /move addresses its git stash precisely so a concurrent stash can't misplace changes, and worktree creation skips a leftover directory instead of failing
• Show the exit resume hint for a renamed session with no messages when its workspace is only available as a cached snapshot
• Show a connected message after a slow MCP server eventually connects
• Add update/uninstall verbs to /plugins, let enable/disable/remove target plugins, MCP servers, or skills via --plugin/--mcp/--skill flags or a positional kind, and support installing skills with /plugins install --skill
• Add a /plugins help command plus skill, MCP, and marketplace management for full /plugin parity
• Session exports keep angle brackets intact in inline code and top-level fenced code blocks
• Show MCP server status correctly for names like constructor and proto
• Keep the session highlight on the nearest live row when closing a session, in both the Sessions split view and the standalone Sessions tab
• Pad inline hex color swatches once inside Markdown lists
• Type $ at the prompt to open an interactive shell in the current session directory (enable it with /settings shellShortcut on; off by default)
• Nested markdown lists render correctly in buffered output (-p --stream off and detail screens): sub-bullets are no longer glued onto the parent item's line or flattened, and are indented under their parent
• copilot skill list now strips terminal control characters from skill names and descriptions, so a crafted skill can no longer inject ANSI escape sequences into the listing output.
• Install skills from the CLI with copilot plugins install --skill <file, URL, or directory> (add --scope project to a file or URL install to install into the repository)
• Show default values in /settings and let booleans cycle back to default
• Require SSO for remote control when managed settings demand it
• Mask secret values in /settings show output
• Show hex color codes written as inline code (e.g. #FF0000) as color swatches, and add a renderHexColors setting (on by default) to toggle hex-color swatches
• Add /model --session (-s) to change the model, reasoning effort, or context window for just the current session, leaving global settings unchanged.
• Detect VS Code, Cursor, and Windsurf through parent processes in /terminal-setup
• The Sessions sidebar is now navigable with the keyboard and mouse (arrows open and focus it and move the selection, and Enter or a click switches to a session; press n to spawn a session or x twice to close one from the keyboard); /settings can disable it or stop restoring remembered sessions
• Add --plugin, --mcp, and --skill flags for plugin mutations
• Add skill removal support to copilot plugins remove --skill
• Wrap ask-user and elicitation inputs in the split-pane chat view
• Modified vim keys (Ctrl+K, uppercase J/K) no longer move the selection in tool-permission prompts and other text-input select menus; only unmodified j/k, the arrow keys, and Ctrl+P/Ctrl+N navigate.
• /terminal-setup now refuses to modify a VS Code keybindings.json that contains a JSON syntax error (instead of rewriting it and reporting success), matching its documented invalid-JSON handling.
• Reveal full file paths when expanding compact editing rows
• Make the plan-approval menu deterministic across models
• Keep /add-dir directories visible in the agent context across turns
• Multi-turn subagents are always enabled, so you can send follow-up messages to running agents
• Enable tool search for Claude Haiku 4.5+
• Emoji shortcodes like 🎉 no longer render with a spurious trailing space in printed and PR/issue/gist output
• Deliver scheduled prompts as steering messages when the agent is busy

1.0.72-1

Added

• Add --plugin, --mcp, and --skill flags for plugin mutations
• Add skill removal support to copilot plugins remove --skill

Improved

• Reveal full file paths when expanding compact editing rows
• Make the plan-approval menu deterministic across models
• Keep /add-dir directories visible in the agent context across turns

Fixed

• Wrap ask-user and elicitation inputs in the split-pane chat view
• Modified vim keys (Ctrl+K, uppercase J/K) no longer move the selection in tool-permission prompts and other text-input select menus; only unmodified j/k, the arrow keys, and Ctrl+P/Ctrl+N navigate.
• /terminal-setup now refuses to modify a VS Code keybindings.json that contains a JSON syntax error (instead of rewriting it and reporting success), matching its documented invalid-JSON handling.

1.0.72-0

Added

• Multi-turn subagents are always enabled, so you can send follow-up messages to running agents
• Enable tool search for Claude Haiku 4.5+

Improved

• Deliver scheduled prompts as steering messages when the agent is busy

Fixed

• Emoji shortcodes like 🎉 no longer render with a spurious trailing space in printed and PR/issue/gist output

1.0.71

2026-07-16

• copilot -p --autopilot no longer hangs when a background shell or agent outlives the turn; it now honors the COPILOT_TASK_WAIT_TIMEOUT_SECONDS timeout the same way plain -p does.
• Reopening the /subagents model picker keeps each agent's reasoning effort and context tier
• Refresh memory context after 30 minutes in long-lived sessions
• Keep MCP tool lists up to date when servers change
• Avoid leaving long-running background git processes after exit
• Add a configurable maximum for Ctrl+R command history
• On startup, an invalid settings.json now shows a warning identifying the offending value instead of silently ignoring your settings
• /terminal-setup no longer skips setup on terminals without real kitty keyboard support
• Add /voice devices to choose and persist the microphone for voice mode
• Limit which built-in agents are available to tasks and subagents
• Add canvas support in the CLI for extension-driven interactions
• Enforce the sandbox filesystem policy on LSP file reads and rename edits
• Allow empty owner and author emails in marketplace metadata
• Keep all MCP Server Type options visible on short terminals
• Mark disabled skills in copilot skill list and its JSON output
• Plan mode now hard-blocks built-in tool calls that would modify the workspace, so the agent can no longer edit files or run mutating shell commands while planning (built-in mutators like opening a pull request are blocked; MCP and external tools are still allowed)
• Improve /chronicle cost-tips recommendations with richer cost profiles
• Highlight standalone hex colors inline in Markdown
• Persist GitHub MCP toolset/tool config via settings.json (githubMcpToolsets, githubMcpTools, etc.)
• Add plugins marketplace subcommands to list, add, and remove plugin marketplaces
• Persist sidebar sessions across restarts
• Add plugins marketplace browse and update commands
• Split /worktree and /move: /worktree now creates a new worktree and leaves your uncommitted changes behind, while the new /move carries them into the new worktree
• Add local and cloud cost profiles to /chronicle cost-tips
• Switching to autopilot mid-turn now auto-answers questions asked during that same turn
• Custom agents that request a shell tool by alias now also receive the matching read, list, and stop shell tools
• Slash commands and their autocomplete now match regardless of case (e.g. /SESSION works like /session)
• Show repo-enabled plugins in /plugin list and skill pickers
• Press ? twice to dismiss quick help and start a prompt with a literal ?
• Shell completions suggest positional-argument choices
• Show the /app launch message and download link immediately on Linux
• Validate --max-autopilot-continues rejects NaN, negative, and fractional values
• Honor NO_COLOR in the CLI even when chalk cached a color level
• Apply updated session options (shell flags, streaming, custom agent defaults) immediately after /settings changes
• Announce the focused /model row for screen readers
• Announce the focused picker row to screen readers
• Show selected custom agents once in /agent and keep their source label when the file name differs from the display name
• Clear the /model pricing banner when no models match
• Keep /share file session and /share html session from using the full-session selector as an output path
• Honor --context in fresh interactive sessions
• Fixed the model picker changing a hidden model's reasoning effort or context window when the search matched no results, and hid the inert key hints shown in that empty state
• Display plugin root skills as /plugin instead of /plugin:plugin
• Keep valid hooks in a config file when one hook entry is malformed
• Denying write(path) now blocks only the specified path
• Using --add-github-mcp-tool "*" now enables all GitHub MCP tools
• Render empty untracked files without a phantom added line
• Show clean failure messages when copilot skill add fails
• Press Enter on a blank settings array item to show an error instead of saving an empty value
• Press Enter once to toggle booleans with a registered default in /settings
• Declining folder trust in /cwd keeps your live session open and returns to the previous folder
• Show a warning when a workspace MCP config is malformed or cannot be loaded
• Make bare copilot mcp and copilot skill print help and exit 0, matching copilot plugin. Consistent with plugin, the implicit help <subcommand> form is not supported for these groups; use copilot mcp <subcommand> --help (or copilot skill <subcommand> --help) instead.
• Show malformed allowed_models.txt policy errors cleanly in -p mode
• Resume synced sessions by name without a false multiple matches error
• Show an error when --name is used with --session-id for an existing session
• Show --plugin-dir plugins in copilot plugin list
• Keep backgrounded sessions alive when you switch away from them
• Link bare #number GitHub refs in -p --stream off output
• Show the startup banner only on the first launch when set to once
• Allow copilot update and /update to accept stable as a channel
• Surface --plugin-dir warnings in the terminal
• Surface the real load error for malformed custom agents
• Reject --continue when used with --resume
• Prompt mode now exits non-zero when a --share or --share-gist export fails
• Server mode reconnects OAuth MCP servers from cached tokens
• Keep stored Git credential helpers available for marketplace plugin installs
• The /model picker shows the Auto model description as markdown with a clickable Learn More link
• Keep sessions tied to their working directory across prompts, restarts, and workspace tools
• Always offer a custom answer in ask_user choice prompts
• Lower the default maximum sub-agent nesting depth from 6 to 4 to curb runaway recursive sub-agent delegation. Usage-based billing users can still adjust subagents.maxDepth (up to 128).
• Add a pinned prompts setting in /settings to control prompt pinning
• Add Repo and Repo (local) scope tabs to the /settings dashboard
• Interactive shell commands now fail with a retryable "reconfiguring" message instead of an "unknown shellManager handle" error after the shell manager is disposed, and a detached command's completion notification is no longer lost when the shell context is reconfigured while a read is in progress
• Reject custom-agent names that would create hidden files
• Reject malformed --allow-tool and --deny-tool patterns with an error message
• Show retained shell output in /tasks Shell Details for finished tasks
• Remove duplicate Error: prefixes from plugin command failures
• Shell completions stop suggesting subcommands as flag values
• Show singular message counts in /usage activity graph
• Keep /cd from switching to files or inaccessible directories
• Dismissing the quick-help overlay with ? no longer leaves a stray ? in the prompt
• --sandbox and --no-sandbox now show their "ignored" warning during interactive startup when the sandbox feature is unavailable (previously it was only visible in non-interactive mode)
• Show the full command with its arguments (not just the wrapper) in the /mcp server detail view
• Hide the inert navigate and view-log hints in the empty /lsp logs (LSP Services) panel
• Exit non-interactive prompt runs with a failure code when a prompt is blocked before responding
• Show the Auto discount in the redesigned inline model picker
• New sessions start in the default directory instead of the active session's cwd
• Fish completion only offers enum values for closed-choice flags
• Use targeted validation commands and lighter install guidance by default
• Use ctrl+x → x to close a session and ctrl+x → h to hide the split sidebar