ADO Code

AI coding assistant with Azure DevOps work item integration for VS Code.

Features

• Fetch work items from Azure DevOps in one tree view with a mode toggle — My Work Items / All Work Items / Unassigned — rendered as a real hierarchy (Epic → Feature → User Story → Task)
• AI chat assistant with OpenAI-compatible and Anthropic-compatible LLM support
• Tool calling with agentic loop for autonomous coding tasks (Chat/Plan/Act/YOLO modes) — independent tool calls run in parallel (even in mixed batches: consent-free calls don't wait behind approval cards), batched edit_file edits, batch get_work_item ids / run_terminal_command commands arrays, and a search_files grep tool with per-line truncation
• Git-based task workflow: auto-create branch on pickup, update CHANGELOG.md on completion
• External agent orchestration (Claude Code, Codex, OpenCode, Hermes, Pi, and more)
• Git worktree isolation: Concurrent agents run in isolated worktrees — no branch conflicts
• Worktrees view: Dedicated sidebar tree with per-worktree details — run status, agent, dirty/clean files, last commit, ahead/behind — plus Open in Terminal/Explorer, Remove, and Show Agent Output actions
• Reopen agent output: Finished runs keep a "↗ Reopen" button and right-click → Show Agent Output — the summary panel is never permanently lost after being closed
• Memory-driven agent hooks: Workspace memory keys agent.before / agent.after run shell commands around every agent invocation; user + workspace memory are injected into every agent handoff prompt
• Model capability checking: Image paste/attach is disabled for non-vision models, models without tool calling are flagged, and the Configuration page shows a live capability readout. Detection is three layers — per-model overrides (you declare it, it wins) > live gateway data (OpenRouter/Ollama capability fields on /models) > name heuristic (gpt-4o/5, claude, gemini, o1–o9, llama-4, …)
• Capability overrides: adoCode.llm.capabilityOverrides declares vision/tool support per model id when auto-detection gets it wrong — edited as structured rows in the Configuration page, beating every other detection layer
• Multi-agent output panels: track multiple agent runs simultaneously with independent streaming output
• File checkpoints: auto-save before AI edits, restore on demand
• MCP (Model Context Protocol) support for external tool servers
• Slash commands with autocomplete (/help, /status, /clear, /mode, /delegate, …)
• User memory: AI remembers your preferences across sessions
• Workspace memory: project-specific conventions in .ado-code/memory/ — kept out of git/Docker via auto-ignore
• .ado-code auto-ignore: The workspace data directory is offered to be added to .gitignore/.dockerignore on load (setting adoCode.ignore.dotAdoCode, default on)
• Direct mode selection: click Chat/Plan/Act/YOLO to switch modes instantly
• In-chat confirmation cards: task start, mode switch, and other prompts render as styled cards inside the chat
• AI choice detection: When the AI asks you to choose, options appear as clickable buttons — clicking one sends the option as a follow-up message; natural-language offers ("Want me to …?") are parsed via an optional cheap model (adoCode.llm.choiceDetectionModel), and long options stack as full-width rows
• Reorganized Configuration UI: Settings are grouped into navigable sidebar categories — Connection, AI & Modes, Permissions, Workflow, and Integrations — each with a setting-count badge, so you can jump straight to what you need instead of scrolling. The Advanced toggle gates a dedicated Advanced category: per-mode model selection, capability overrides, the choice detection model, and native token counting
• Project Creation Wizard: Multi-step UI for creating new projects with 9 templates (Node.js, Python, React, Next.js, Laravel, .NET) — configure options, ADO integration, and git setup before creation (/new-project)
• Wizard focus: The Configuration page, project-creation wizard, and first-run setup take over the whole sidebar while open — the sibling views (Work Items / Status / Worktrees) collapse automatically and are restored when the wizard closes, so wizards never share space with the trees. The chat kebab's "Configuration…" opens the in-app Configuration page
• Skill Management System: Browse, install, and manage reusable AI skills — 10 built-in skills (Code Review, Documentation, Testing, Refactoring, Security Audit, Performance Profiler, Deployment Checklist, Database Schema Review, Accessibility Audit) with search, filtering, and enable/disable toggle (/skills)
• AI skill execution: The AI can discover and use enabled skills during conversations via the execute_skill tool — "Execute in Chat" button injects the skill prompt and triggers the LLM
• Skill import: Import custom skills from local .json files, SKILL.md files (YAML frontmatter + markdown), or .tar.gz/.zip archives — imported skills persist across VS Code restarts
• External skill registries: Browse and install community skills from remote TSV registries (slug<TAB>url<TAB>description) — the built-in UI Skills registry is always included, and you can add more via adoCode.skillRegistryUrls. Registry skills carry a "registry" badge in the Skill Catalog and are downloaded + imported (SKILL.md) when you install them
• Send Selection to Chat: Right-click selected text in the editor → "ADO Code: Send Selection to Chat" inserts it as a fenced code block in the chat draft
• Send File to Chat: Right-click a file in the Explorer → "ADO Code: Send File to Chat" attaches it as a chip above the input bar with full content ready to send
• Delete All Sessions: /clear-sessions slash command + "🗑️ Delete All Sessions" in session history and kebab menu — wipes all sessions after in-chat confirmation
• Confirmation cards for destructive operations: Single session delete and delete-all show in-chat confirmation cards with danger styling instead of silently executing
• Categorized /help: Help output is grouped into Work Items, Chat, AI, and Other sections with autocomplete tip
• AI merge flow: For finished agent runs, the assistant can commit, push, and open an ADO pull request itself (commit_worktree → push_worktree → create_pull_request) — never force-pushes, never touches protected branches (adoCode.git.protectedBranches, default main/master), refuses to commit failed runs unless it explicitly overrides after reviewing, and can resolve merge conflicts (resolve_pr_conflicts) then re-commit/re-push until the PR is clean
• Dynamic context window detection: Context window size auto-detected from the /models API endpoint (Ollama n_ctx, OpenRouter context_length, etc.) — live data overrides the hardcoded table; content-aware token counting (code ~3.5, prose ~4.5 chars/token) with expanded model table (27 models) and 128k default
• Context management: Priority-based conversation truncation replaces naive 20-turn cutoff; real token counter replaces rough char/4 estimation; conversation auto-condenses at 75% context via LLM summarization; token status bar shows accurate model-aware counts
• Mode-aware system prompt: Dynamic prompt generation includes mode-specific role, available tools, tool guidelines, and environment context
• MCP server management: Right-click MCP servers in the Status Panel to disconnect, reconnect, or view details Mode quick-cycle: Right-click the Mode item in the Status Panel to cycle through inline/plan/act/yolo
• Worktree batch cleanup: Remove All Completed action on the Worktrees root node
• Agent run history: Recent Runs section in the Status Panel shows last 10 completed runs with status icons, timestamps, and context menus
• Agent details panel: Right-click agent in Status Panel to view name, binary, version, supported modes, and CLI arguments
• Memory search: QuickPick fuzzy search across all user and workspace memory entries
• Memory import/export: Export memories to JSON; import with merge or replace option
• Work item filtering: Filter Work Items tree by state, type, or text search with smart parent visibility
• Agent auto-review: Git diff automatically reviewed by LLM on agent completion with merge recommendation
• Worktree diff viewer: Show Changes opens VS Code diff editor for worktree files
• Generate tasks from user stories: /generate-tasks slash command + context menu action trigger the AI to analyze a user story and create child tasks via the create_work_item LLM tool — with duplicate prevention (checks for existing child items before creating)
• Task draft editor: When the AI calls create_work_item, an editable markdown tab opens for you to review and modify all fields before the work item is created in ADO — markdown content is automatically converted to HTML for proper rendering in ADO
• Parent delegation completes child items: Delegating a parent work item (Story/Feature/Epic) covers its entire descendant subtree — the agent gets a numbered delivery checklist of every open child and must end with a ## Delivery Report marking each DONE / BLOCKED / INCOMPLETE. With adoCode.agents.autoCompleteChildren enabled, the extension then comments + transitions each done child to its terminal state and closes the parent once all children are done (partial success leaves it open with a comment). Without the setting, the run reports and comments only. You're warned before delegating and can skip the children; a parent/child concurrency guard prevents overlapping runs
• Agent progress in chat: Agent delegation shows start/completion messages in the chat thread alongside the streaming output panel — with a live elapsed-time readout while a run is in flight
• Live agent progress in the editor: Move any agent run's progress to the editor area — real-time streaming output with timestamps and event styling, live elapsed clock, progress bar, and the final summary in the same panel. Auto-open per run via adoCode.agents.progressView = editor, or on demand via the ↗ button on the chat output panel (running or finished) or the ADO Code: Open Agent Progress… command
• Changelog notification on update: After a version change, a one-time notification offers to show the new version's changelog in a styled webview panel
• Keyboard shortcuts: Ctrl+Shift+M (cycle mode), Ctrl+Shift+/ (search memories), Ctrl+Alt+R (refresh status)
• Clean Up After Merge: Removes a merged run's worktree and deletes its branch (only when the PR is actually merged and the branch is fully merged) — from the Worktrees view context menu or an auto-offer after a merged PR
• Working indicator: A status-bar spinner shows while any LLM turn or agent run is in flight — visible even when the chat view is hidden, with live detail ("thinking…", "tool: edit_file")
• Full detail view: right-click → "Show Full Details" opens a formatted panel in the editor
• Work item selection: Selected items show a green badge; deselect via the row context menu
• Agent delegation injects AGENTS.md project context so all agents understand the codebase
• Agent summary panel: Completion summaries open in a styled HTML panel with metadata and duration
• Workspace scaffolding: new project wizard for Node.js, Python, PHP (Laravel), .NET C# with auto-generated files
• Workspace-to-ADO project binding: prevents working on items from the wrong project

Requirements

• VS Code 1.85.0+
• Node.js 18+
• A git-enabled workspace folder
• Azure DevOps PAT (Personal Access Token) with vso.work_write scope
• An LLM API key (OpenAI-compatible or Anthropic)

Extension Settings

Configure in VS Code settings under adoCode.*. The Configuration page organizes these into sidebar categories, each with a setting-count badge:

• Connection — Azure DevOps + LLM Provider
• AI & Modes — Mode, Act Mode, Chat, Sessions
• Permissions — Consent
• Workflow — Git, Changelog, Work Items, Workspace
• Integrations — Agents + MCP Servers
• Advanced — gated by the Advanced Configuration toggle; holds per-mode model selection, capability overrides, choice detection model, and native token counting (toggling it on jumps straight there)

The Advanced Configuration toggle in the sidebar enables the Advanced category for per-mode model selection and reasoning effort tuning.

Setting	Description
adoCode.organizations	ADO organizations (array of { name, url, project })
adoCode.adoOrganization	Active Azure DevOps organization name
adoCode.adoProject	Active Azure DevOps project name
adoCode.adoServerUrl	Optional on-prem ADO Server (TFS) base URL — empty = cloud
adoCode.adoPat	Personal Access Token
adoCode.ado.clarificationState	State set when clarification is requested
adoCode.ado.warnOnSparseTask	Warn before starting tasks with no description/AC
adoCode.llmProvider	openai (default) or anthropic
adoCode.llmApiUrl	LLM API endpoint
adoCode.llmApiKey	LLM API key
adoCode.llmModel	LLM model name (used for all modes in Express mode)
adoCode.llm.choiceDetectionModel	Optional cheaper model for AI choice-prompt detection (empty = main model, off = regex-only)
adoCode.llm.capabilityOverrides	Per-model capability overrides (array of { model, vision?, tools? }) — beats auto-detection
adoCode.advancedConfig	Enable Advanced Configuration mode for per-mode model selection (default false)
adoCode.llm.modeConfigs	Per-mode model overrides in Advanced mode: { "inline": { "model": "gpt-4o" }, "act": { "model": "o3" } }
adoCode.llm.modeReasoningEffort	Per-mode reasoning effort for reasoning models: { "inline": "low", "act": "high" } — values: low, medium, high
adoCode.llm.useNativeTokenCounting	Use provider-native token counting (Anthropic count_tokens; OpenAI-compatible usage.prompt_tokens) for status-bar accuracy (default true)
adoCode.mode	Tool-use mode: inline, plan, act, or yolo
adoCode.act.toolBudget	Max agentic loop iterations per chat turn (each iteration = one model round-trip that may run several tool calls in parallel; applies to all modes)
adoCode.act.terminalAllowlist	Allowed command prefixes in act mode
adoCode.consent.harmlessAutoApprove	Auto-approve harmless (read-only) terminal commands after a timer (default false)
adoCode.consent.harmlessAutoApproveSeconds	Seconds before a harmless command auto-approves (default 20, range 1–30)
adoCode.consent.autoApproveTools	Tool names or glob patterns (read_*, get_*, edit_file) that skip the consent prompt in inline/act modes (default [])
adoCode.chat.showThinking	Show the model's thinking/reasoning text while it processes (default true)
adoCode.chat.showToolCalls	Show live tool-call cards in the chat while the AI works (default true; off shows only a pulsing "…" indicator)
adoCode.git.requireGitRepo	Block task pickup outside a git repo
adoCode.git.createBranchOnTaskStart	Auto-create feature/ADO-<id>-<slug> branch
adoCode.git.requireCleanTree	Warn on branch switch with uncommitted changes
adoCode.git.prOnCompletion	Offer to push + create a PR via gh on task done
adoCode.git.protectedBranches	Branches create_pull_request refuses to target (default ["main", "master"])
adoCode.changelog.enabled	Update CHANGELOG.md on task completion
adoCode.changelog.autoCommit	Commit CHANGELOG.md automatically
adoCode.changelog.postToAdo	Post the changelog entry as an ADO comment
adoCode.ignore.dotAdoCode	Offer to add .ado-code to .gitignore/.dockerignore (default true)
adoCode.sessions.maxPerProject	Max chat sessions kept per workspace + project
adoCode.agents.enabled	Which external agents may be delegated to
adoCode.agents.verifyCommand	Shell command run after an agent finishes
adoCode.agents.autoSelect	Default agent when none is specified
adoCode.agents.autoReview	Auto-review agent changes via LLM when a run completes (default true)
adoCode.agents.autoCompleteChildren	After a delegated parent run succeeds, auto-complete its children in ADO: children marked DONE in the agent's Delivery Report transition to their terminal state (Task/Bug → Closed, Story/Feature/Epic → Resolved), and the parent closes when all open children are done (default false; off = report + comment only)
adoCode.mcp.servers	MCP server configurations (array of { name, command, args?, env?, timeout? })
adoCode.skillRegistryUrls	Extra skill registry URLs to browse in the Skill Catalog (TSV format: slug<TAB>url<TAB>description); the built-in UI Skills registry is always included

Slash Commands

Type / in the chat input to see available commands:

Command	Description
/status <state>	Set work item state (e.g. Active, Done, Closed, Removed)
/comment <text>	Post a comment to the active work item discussion thread
/pick	Browse and select a work item from the tree to set as active context
/assign <who>	Assign the active work item to a team member (name or email)
/clear	Clear all chat messages and start fresh
/clear-sessions	Delete all chat sessions and start completely fresh
/mode [mode]	Switch tool mode: inline (ask), plan (read-only), act (auto-approve), or yolo (full autonomy) — omit mode to pick from list
/undo	Revert the last state change made to the active work item
/help	List all available slash commands with usage examples
/delegate [agent] <prompt>	Hand off the active task to an external agent (claude, codex, opencode, hermes, pi, gemini, dsh)
/generate-tasks	Generate child tasks for the active user story (review in editor before saving)
/new-project	Open the project creation wizard to create a new project
/skills	Open the skill catalog to browse and manage AI skills
/resume	Switch to a previous chat session to continue where you left off
/remember <what>	Store a preference or instruction the AI will remember across sessions
/forget	Remove all saved notes and preferences

Memory System

User Memory

AI remembers your preferences across all sessions:

• "Remember that I prefer conventional commits"
• "Don't use var, only let/const"
• Use /remember and /forget commands

Workspace Memory

Project-specific conventions in .ado-code/memory/:

• conventions.md — Coding style, naming patterns
• architecture.md — Design decisions, rationale
• gotchas.md — Known pitfalls, workarounds
• custom.md — Anything else project-specific
• agent.before.md / agent.after.md — Agent hooks: shell commands run automatically before/after every agent invocation (e.g. npm ci && npm run lint before, npm test after). Output streams to the run panel / is captured into the summary

Memory is injected into the chat assistant's system prompt and into every external-agent handoff prompt ("ADO Code Memory (instructions you MUST honor)"); the executable hook keys are excluded so agents don't re-run them.

.ado-code/ is per-user, machine-local state — the extension offers to add it to .gitignore and .dockerignore on load (see adoCode.ignore.dotAdoCode). To share conventions with your team, prefer documenting them in AGENTS.md.

Agent Worktrees

Each delegated agent run gets an isolated git worktree under .ado-code/worktrees/:

• The Worktrees view (ADO Code activity bar) lists every worktree with its run status (running/succeeded/failed/interrupted), branch, dirty/clean files, last commit, and ahead/behind
• Right-click actions: Open in Terminal, Open in Explorer, Commit & Push, Remove, Show Agent Output
• Branches are named from the ADO work item subject, e.g. feature/ADO-42-fix-login-bug
• Merge flow: after a run finishes, commit its changes (Commit & Push or the AI's commit_worktree/push_worktree tools), then open a PR (create_pull_request) — the AI is instructed to execute this flow and will stop and report if a step fails
• Parent runs: delegating a parent work item runs the whole subtree in ONE worktree/branch/PR — children are a delivery checklist inside the prompt, and the post-run sync (see the settings table) updates ADO states when it finishes
• Guardrails: two agents can never run on the same work item concurrently — including overlapping parent/child runs (a parent run covers its descendants); reusing a stale branch warns; Remove refuses silently discarding dirty work (offers Commit & Push, then Remove) and deletes the local branch afterwards only when it is fully merged — commits are never lost

Model Capabilities

The extension infers the active model's capabilities from its id:

• Vision: models without image support get the chat's image paste/attach disabled
• Tool calling: models without it show a persistent warning that agentic modes degrade to plain chat
• The Configuration page shows a live readout ("Capabilities: vision · tool calling") for the selected model and highlights models lacking tool calling

Release Notes

Unreleased

• Mixed-batch tool calls stay parallel: When a model turn mixes consent-free calls (reads, allowlisted commands) with calls that need your approval, the free ones now run concurrently instead of the whole batch serializing behind the first consent card — approval cards still appear one at a time, never stacked
• Batch work-item reads: get_work_item accepts an ids array — several work items in one call (up to 20, batch-fetched with discussion threads loaded in parallel)
• Batch terminal commands: run_terminal_command accepts a commands array — several commands in one call, one consent card, outputs concatenated; every command is still checked for shell operators and the act-mode allowlist
• Wizards take over the sidebar: Configuration page, project-creation wizard, and first-run setup collapse the sibling views (Work Items / Status / Worktrees) while open and restore them on close
• "Configuration…" opens the in-app Configuration page: The chat kebab no longer jumps to VS Code's native settings — it opens the in-webview Configuration page (full-width, with the sibling views collapsed)

0.6.3

• New Project wizard actually creates projects: The wizard always failed with "No target path specified" — the webview sent an empty target folder for the host to resolve and the host never did. The host now resolves the destination itself: your open workspace folder (blank directories included) is used automatically, or you're asked to pick a folder when none is open
• Empty-workspace prompt opens the full wizard: The "Create New Project" offer in an empty folder previously used a limited inline flow (6 templates) — it now opens the full wizard with all 9 templates and every step
• Wizard template defaults apply when you skip the options step: Picking a template pre-fills its option defaults (README/.gitignore/LICENSE for Empty, ESLint/Prettier/Jest for Node, Tailwind for Next.js, …)
• Wizard git branch name is honored: "Initial Branch Name" was collected but ignored — the chosen branch (default main) is now actually created
• ADO integration creates the work item: Enabling it now creates the work item in your active project (type + area path included); failures show in the success toast and never block creation
• Wizard errors are visible again: Failures render inside the wizard (the old global banner was hidden behind the overlay) and Create re-enables after a failure
• Scaffolded projects are more solid: Laravel gets the missing bootstrap/app.php and valid PHP namespaces; composer.json PSR-4 autoloads correctly; the Node Dockerfile uses npm install; .NET solutions get a real project GUID; Next.js Prisma/Tailwind options actually take effect
• Project-name validation: Empty or path-traversing names are rejected up front
• 24 new tests: every one of the 9 templates is exercised end-to-end (scaffold, git init, commit, branch) plus options, validation, and the blank-directory flow

0.6.2

• Images inside work items now display: Rich-text fields (Description, Acceptance Criteria, Repro Steps, System Info) and discussion comments can contain <img> tags pointing at ADO's attachment endpoint — those URLs need authentication a webview can't attach, so they used to render as broken images. The extension now fetches each attachment with your PAT and inlines it as a data:image/... URL in the chat task panel and the standalone work-item panel (capped at 10 images / 2 MB each / 8 MB total per field)
• Work Items view shows assignment at a glance: Every tree row now carries an assignment cue — blue person = assigned to you, orange person = assigned to someone else, grey empty circle = unassigned — with the work item type moved into the row description (#123 · Task) and the assignee in the hover tooltip
• Generated tasks post reliably: The generate-tasks flow only understood a ```json fence — if the model answered with a numbered list ("1. … 2. …") or a ```JSON/bare-fence/unfenced array, the whole batch was silently dropped. The parser now tolerates all of those (and falls back to a numbered markdown list), so every proposed task reaches the review editor
• Numbered sub-items inside task descriptions survive posting: Task descriptions/acceptance criteria were truncated to their first line when the review file was parsed back; multi-line bodies with numbered/bulleted steps are now captured in full
• Changelog comments post cleanly formatted: The completion-flow ADO comment is now block-structured (ADO collapses single newlines) with a clickable "Open in Azure DevOps" link and a backticked branch/commit bullet; work-item URLs are percent-encoded so project names with spaces no longer break the link
• Created work items render descriptions correctly: markdownToHtml was rewritten as a line-based parser — mixed bullet/numbered lists now produce valid <ul>/<ol> HTML (the old greedy regex left bare <li>s ADO rendered as plain text)

0.6.1

• Proper tree structure for the Work Items tree: The tree now renders the real ADO hierarchy instead of flat roots. The extension walks up the parent chain (Task → User Story → Feature → Epic) and down the children, so a Feature expands to show its User Stories, and Tasks nest under their parent Story/Task even when those parents are assigned to someone else. Hierarchy-context items are tagged "· context" (with an explanatory tooltip); if the child query is rejected by your ADO org it falls back to per-parent queries, and if expansion still fails the tree falls back gracefully with a visible warning
• Merged Work Items view with mode toggle: My Work Items, All Work Items, and Unassigned Work Items are now ONE tree view ("Work Items") — a visible header row at the top of the tree ("View: My Work Items — click to switch ▾") shows the current mode and opens the picker, just like the Chat|Plan|Act|Yolo toggle. All mode fetches every open item in the project; the choice is remembered per workspace, and context menus are gated per item so Take Ownership / Reassign appear on unassigned items in every mode
• Tolerant System.Parent parsing: some ADO orgs return the parent link as a bare number instead of { id } — both shapes are parsed, so the tree nests regardless of your org's serialization
• Tree diagnostics: Output → ADO Code now logs the expansion counts, a raw System.Parent sample (once), and per-refresh tree shape (N items (K with parentId, R roots)) — a flat tree is instantly diagnosable
• Parallel tool execution: Independent tool calls in one turn now run concurrently and results are re-ordered back to call order — several read_file/search_files calls finish as fast as one. Batches containing a consent card run sequentially so prompts never stack. Same-file edits are serialized by a per-file mutation queue so parallel batches can't race
• Batched edits: edit_file accepts an edits[] array — several disjoint changes to a file in a single call (each verified; applied in order)
• search_files grep tool: Now actually implemented — regex search across the workspace with path:line hits, 500-char per-line truncation, workspace confinement, sensible exclusions, and result caps
• execute_skill fixed: The AI can now really call execute_skill to load skill instructions mid-conversation (previously it errored as an unknown tool)
• Truncated-response guard: If a response hits the output token limit, tool calls are failed with "re-issue" instead of executing possibly-truncated arguments
• Architecture cleanup: Removed the legacy BaseTool/ToolRegistry/definitions layer and the abandoned BaseProvider (-v2) migration — tool execution now lives in one switch; −2,500+ lines
• Parent delegation completes child items: Delegating a parent work item now covers its whole descendant subtree (children, grandchildren, …). The agent prompt carries a numbered delivery checklist of every open descendant, and the agent must end with a ## Delivery Report marking each DONE / BLOCKED / INCOMPLETE. With the new opt-in adoCode.agents.autoCompleteChildren setting, the extension then comments + transitions each done child to its terminal state (Task/Bug → Closed, Story/Feature/Epic → Resolved) and closes the parent once all open children are done — partial success leaves it open with a comment, and failed runs never auto-close anything. A parent/child concurrency guard prevents overlapping runs on the same subtree
• Main-model choice offers: When the model ends its answer by offering you a choice, it appends a fenced ```choice block that the host parses as the primary path (no regex misses, no extra LLM call) and the chat renders as a clickable option card — regex and cheaper-model detection remain as fallbacks
• Active editor context auto-injected: The file you're editing is included in each chat turn's context automatically (deduped by file + version; oversized selections capped)
• Slimmer session persistence: sessions persist only the last 25 message pairs plus the condensation marker
• AGENTS.md honored in chat: the chat model reads and honors AGENTS.md when present (on-demand, no cost when absent)
• Conversation summaries fixed: condensation summaries are now marker-prefixed user messages — previously mid-array system messages were silently dropped by the Anthropic providers
• Laravel scaffold fixed: generated artisan/routes/web.php previously lost PHP namespace separators (\C escapes collapse in JS template literals) — they now emit valid PHP
• Lint-clean + dead code removed: 0 ESLint errors; unused ContextProxy module, two unused webview components, and a dead helper deleted; tsconfig now enforces noUnusedLocals/noUnusedParameters
• External skill registries: browse and install community skills from remote TSV registries (built-in UI Skills registry + custom adoCode.skillRegistryUrls); registry skills get a "registry" badge in the Skill Catalog and are imported on install
• Security: dompurify bumped to 3.4.14 in the webview bundle
• dsh delegation for existing users: a stale adoCode.agents.enabled (the pre-dsh default) no longer hides DeepSeek Harness — the registry migrates it automatically, so dsh is delegatable again; custom pruned lists are respected

0.5.9

• Reorganized Configuration UI: Settings are now grouped into navigable sidebar categories — Connection, AI & Modes, Permissions, Workflow, and Integrations — each with a setting-count badge, so you can jump straight to what you need instead of scrolling through everything. The Advanced toggle now gates a dedicated Advanced category: per-mode model selection, Model Capability Overrides (previously settings.json-only), the choice detection model, and native token counting are all editable in the UI — toggling it on jumps straight there
• Live thinking + tool progress: The chat window now streams the AI's reasoning into a 💭 Thinking block and shows each tool call as a live card that appears as running… (spinner, arguments visible) and flips to completed/error with its result when done — clear progress instead of waiting through a silent "Thinking…" for the final answer
• Permanent tool-call record: Tool cards are merged into the final assistant message as collapsible blocks above the answer — expand any call to see its arguments and result, and the record survives session history
• Hide tool calls in chat: Configuration → Chat → Show tool calls in chat (default on) hides the cards; tools still run, and chat shows only a subtle pulsing "…" — no tool names, arguments, or results leave the host
• Show AI thinking now enforced: The Show AI thinking toggle (Configuration → Chat) previously had no effect — it now actually shows/hides the reasoning block in every streaming path
• Wildcard permissions: Configuration → Consent → Auto-approve tools (wildcards) accepts patterns like read_*, get_*, or exact names (edit_file) that skip the consent prompt in inline/act modes. Act-mode Terminal allowlist entries also support wildcards: git *, git push *, npm run * — token-by-token matching with no shell operators
• Token consumption optimization: tool results are capped to a token budget and older results are compacted once the model has seen them; plan mode sends only read-only tool schemas; the token budget now counts the system prompt + tools
• Provider-native token counting: the token status bar uses the provider's own tokenizer (Anthropic count_tokens; OpenAI-compatible via usage.prompt_tokens) with the local estimate as fallback — toggle adoCode.llm.useNativeTokenCounting (default on)

0.5.8

• Project Creation Wizard: Multi-step UI for creating new projects with 9 templates (Node.js TypeScript/JavaScript, Python, React, Next.js, PHP Laravel, .NET Web API/Console, Empty) — configure per-template options (ESLint, testing, Docker, etc.), optional ADO work item creation, git initialization with branch naming, review step before creation (/new-project)
• Skill Management System: Browse, install, and manage reusable AI skills — 10 built-in skills (Code Review, Documentation, Testing, Refactoring, Security Audit, Performance Profiler, Deployment Checklist, Database Schema Review, Accessibility Audit) with search, category filtering, and enable/disable toggle (/skills)
• AI skill execution: The AI can discover and use enabled skills during conversations via the execute_skill tool — "Execute in Chat" button injects the skill prompt and triggers the LLM
• Skill import: Import custom skills from local .json files, SKILL.md files (YAML frontmatter + markdown), or .tar.gz/.zip archives — imported skills persist across VS Code restarts
• Express/Advanced Configuration: New configuration mode toggle — Express (default) uses one model for all modes, Advanced enables per-mode model selection and reasoning effort tuning (low/medium/high) for reasoning models like o1/o3
• Send Selection to Chat: Right-click selected text in the editor → "ADO Code: Send Selection to Chat" inserts it as a fenced code block in the chat draft
• Send File to Chat: Right-click a file in the Explorer → "ADO Code: Send File to Chat" attaches it as a chip above the input bar with full content ready to send
• Delete All Sessions: /clear-sessions slash command + "🗑️ Delete All Sessions" in session history and kebab menu — wipes all sessions after in-chat confirmation
• Confirmation cards for destructive operations: Single session delete and delete-all show in-chat confirmation cards with danger styling instead of silently executing
• Categorized /help: Help output is grouped into Work Items, Chat, AI, and Other sections with autocomplete tip

0.5.7

• Task draft editor: AI create_work_item tool opens an editable markdown tab for review before creating work items in ADO
• YOLO mode: fully-autonomous mode that auto-approves every tool including shell commands — no consent prompts, no allowlist
• Consent auto-approve timer: harmless (read-only) terminal commands show a countdown timer and auto-approve on expiry (configurable, default 20s)
• Show AI thinking: models with reasoning tokens (o1/o3, Claude extended thinking) display their internal reasoning in a collapsible block
• Improved context size detection: content-aware token counting (code ~3.5, prose ~4.5 chars/token), expanded model table (27 models), better substring matching, default raised to 128k
• Dynamic context window auto-detection from /models API (Ollama n_ctx, OpenRouter context_length, etc.) — live data overrides hardcoded table
• Context management with priority-based truncation and auto-condensation
• Mode-aware system prompts with tool filtering
• MCP server management (disconnect/reconnect per server)
• Mode quick-cycle from Status Panel
• Worktree batch cleanup (Remove All Completed)
• Agent run history with status icons and context menus
• Agent details panel (view agent info)
• Memory search, import/export
• Work item filtering (by state, type, text)
• Agent auto-review on completion (LLM-powered code review)
• Worktree diff viewer (VS Code diff editor)
• Keyboard shortcuts for common actions
• Generate tasks from user stories (/generate-tasks + context menu)
• Agent progress shown in chat thread (delegation start/completion messages)
• Changelog notification on package update
• Merge-flow guardrails round 2: commit_worktree refuses failed runs unless allowFailed is set; create_pull_request refuses protected branches (adoCode.git.protectedBranches); new resolve_pr_conflicts tool surfaces git merge conflicts (base/our/their contents) for LLM-driven resolution
• Child task delegation: agent delegation now checks for child work items and warns the user before including them in the agent context
• Clean Up After Merge: context-menu action + auto-offer after merged PR removes the run's worktree and deletes its branch (only when the PR is actually merged and the branch is fully merged)
• Capability overrides: adoCode.llm.capabilityOverrides lets you declare vision/tool-calling support per model id, beating every auto-detection layer — structured editor in Configuration page
• Live model capabilities: detection is now three layers — user override > live gateway data (OpenRouter/Ollama /models) > name heuristic (now covers gpt-5, o1–o9, llama-4, gemma-3, deepseek-vl, glm-4.5v)
• Pre-commit hook clears git env vars so test suites don't inherit hook state

0.5.6

• Dedicated Worktrees view with per-worktree details and Show Agent Output
• Reopen agent output from the chat panel or Worktrees view
• Memory-driven agent hooks (agent.before / agent.after) + memory injected into agent handoffs
• Model capability checking (vision/tool-calling) with UI gating and warnings
• Model list retrieval in the Configuration page
• .ado-code auto-ignore for .gitignore/.dockerignore
• Branch names slugged from the ADO work item title
• Dismissed agent runs no longer reappear; summary panels dedupe on reopen
• AI merge flow: commit/push/PR tools for finished agent runs + worktree guardrails (concurrent-run block, stale-base warning, dirty-remove protection, merged-branch cleanup)
• Working indicator in the status bar (survives chat view being hidden)
• Background LLM processing survives panel close; pending consent cards re-post on return
• AI choice detection via optional cheap model; choice buttons now send the option as a message
• Sessions auto-create on the first message
• Full Configuration page coverage (organizations editor, choice-detection model, workspace ignore toggle) — fixes MCP-server wipe on save

0.5.5

• Git worktree isolation for concurrent agents
• Multi-agent output panels
• Agent summary in a formatted editor panel
• Status Panel context menus (memory, agents, worktrees)
• Work item selection highlighting + deselect
• AI choice prompt detection

0.5.4

• In-chat confirmation cards for all prompts
• Hierarchical work item trees
• Full detail view in the editor
• Workspace scaffolding + AGENTS.md generation + git init prompt
• Workspace-to-ADO project binding
• MCP servers configuration page

0.4.0

• Agent reference files (AGENTS.md, playbooks, structure scripts)
• User memory and workspace memory systems
• Slash commands with autocomplete
• Direct mode selection (click to switch)
• Pre-commit quality gates

0.3.0

• Checkpoint system for file save/restore
• MCP integration for external tool servers
• ADO API version update to 7.1 GA

0.2.0

• Core architecture: BaseTool, BaseProvider, ToolRegistry
• 5 file tools: read, search, edit, execute, list
• Modes system: code, architect, ask, debug
• Context management and UI components

0.1.3

• Agent delegation with Claude, Codex, OpenCode, Hermes
• Task detail review and clarification workflow

0.0.1

Initial release.