An Azure DevOps plugin for Claude Code that turns your sprint backlog into working code.
Point it at your sprint, and it will read your stories, analyze the relevant repos, write implementation plans, generate code, create PRs, and resolve tasks — all without leaving your terminal. Run /devsprint-execute and walk away. Come back to pull requests linked to your stories, tasks marked as resolved, and a summary of everything it did.
Your sprint board stays in sync because the plugin updates it as it works.
- View your sprint from the terminal — stories, tasks, state, descriptions, all color-coded
- Create stories and tasks from natural language — just describe what you need
- Analyze stories — reads the target repo, traces code paths, identifies the files that need changes, and writes a detailed implementation spec
- Execute stories autonomously — creates a feature branch, writes the code, runs tests, commits, pushes, creates a PR linked to the story, and resolves all tasks
- Batch mode — run
/devsprint-executewith no arguments and it loops through every story in your sprint. Errors on one story don't block the next - Fix PR comments —
/devsprint-pr-fix <story-id>fetches review comments, fixes them, and pushes
Zero external dependencies. Just Node.js built-ins and the Azure DevOps REST API.
End-to-end example: /devsprint-create → /devsprint-plan → /devsprint-execute → review PR → /devsprint-pr-fix.
> /devsprint-create Add dark mode toggle: 1) Backend user preference 2) Frontend toggle 3) Tests
Creating:
[User Story] "Add dark mode toggle to settings page"
[Task] "Backend: Store dark mode preference in user profile"
[Task] "Frontend: Add toggle switch to settings page"
[Task] "Tests: Dark mode preference persistence and UI toggle"
╔══════════════════════════════════════════════════════╗
║ Story #3040 created ║
╚══════════════════════════════════════════════════════╝
✓ #3040 [User Story] "Add dark mode toggle to settings page"
✓ #3041 [Task] "Backend: Store dark mode preference in user profile"
✓ #3042 [Task] "Frontend: Add toggle switch to settings page"
✓ #3043 [Task] "Tests: Dark mode preference persistence and UI toggle"
Sprint: Sprint 7
Next step: /devsprint-plan 3040 to analyze and create spec.
> /devsprint-plan 3040
#3040 → NorthwindApp (auto-detected)
### #3040 — Add dark mode toggle to settings page (New)
My understanding:
Allow users to switch between light and dark mode from their settings.
The preference is stored in the user profile and applied on login.
Work type: Code change
Target repo: NorthwindApp
Repo analysis:
Tech stack: C# / .NET 8 + React 18 / TypeScript
Key files:
src/Api/Controllers/UserController.cs — user profile endpoints
src/Api/Models/UserProfile.cs — profile entity (add DarkMode bool)
src/Web/pages/Settings.tsx — settings page (add toggle)
src/Web/context/ThemeContext.tsx — theme provider (already exists)
Code flow: UserController.UpdateProfile() → UserService → UserRepository
STORY.md written to NorthwindApp/.planning/stories/3040.md
Changes?
> ok
Planning complete. Run /devsprint-execute 3040 to implement.
> /devsprint-execute 3040
╔══════════════════════════════════════════════════════╗
║ Pre-flight Status Check ║
╚══════════════════════════════════════════════════════╝
Will execute:
→ #3040 — Add dark mode toggle to settings page
State: New | Tasks: 0/3 done | Repo: NorthwindApp
━━━ [1/1] Story #3040 — Add dark mode toggle to settings page ━━━
Baseline tests green — proceeding.
Created branch feature/3040-dark-mode-toggle from develop
Task status updates:
#3041 (Backend: Store dark mode preference): Active ✓
#3042 (Frontend: Add toggle switch): Active ✓
#3043 (Tests: Dark mode persistence and UI): Active ✓
... implementing (TDD: write tests → implement → verify) ...
Task resolution:
#3041 (Backend: Store dark mode preference): Resolved ✓
#3042 (Frontend: Add toggle switch): Resolved ✓
#3043 (Tests: Dark mode persistence and UI): Resolved ✓
Story #3040 resolved ✓
╔══════════════════════════════════════════╗
║ Execution Complete ║
╚══════════════════════════════════════════╝
✓ #3040 — Add dark mode toggle to settings page
Branch: feature/3040-dark-mode-toggle
Tasks: 3/3 resolved
Tests: 22 passed, 0 failed (dotnet test) — all passed
Story: Resolved ✓
PR: https://dev.azure.com/.../pullrequest/187
Review the PR in Azure DevOps. Leave comments on anything that needs fixing, then run /devsprint-pr-fix:
> /devsprint-pr-fix 3040
=== PR #187: #3040 Add dark mode toggle to settings page ===
Status: active
Branch: feature/3040-dark-mode-toggle -> develop
=== Unresolved Review Comments (2) ===
File: src/Api/Models/UserProfile.cs
Line 24: "Should default to false, not null"
File: src/Web/pages/Settings.tsx
Line 89: "Use the existing ToggleSwitch component instead of a raw checkbox"
Fixing all 2 comments...
✓ Fixed UserProfile.cs — changed DarkMode default to false
✓ Fixed Settings.tsx — replaced checkbox with ToggleSwitch component
Tests: 22 passed, 0 failed
=== PR Fix Complete ===
Fixes pushed: 2 commits
PR: https://dev.azure.com/.../pullrequest/187
Run /devsprint-plan and /devsprint-execute without a story ID and they process everything autonomously. Walk away and come back to PRs.
> /devsprint-plan
=== Analysis: Sprint 7 ===
You have 3 stories:
[US] #3040 -- Add dark mode toggle (New) → NorthwindApp
[US] #3044 -- Email notification preferences (New) → NorthwindApp
[US] #3048 -- Fix timezone bug in scheduler (New) → NorthwindApp
Analyzing...
#3040 — already analyzed, keeping existing spec.
#3044 → NorthwindApp (same as other stories)
... analyzing repo, generating spec ...
STORY.md written to NorthwindApp/.planning/stories/3044.md
Changes?
> ok
#3048 → NorthwindApp (same as other stories)
... analyzing repo, generating spec ...
STORY.md written to NorthwindApp/.planning/stories/3048.md
Changes?
> ok
=== Analysis Complete ===
Stories planned:
#3044 Email notification preferences → NorthwindApp/.planning/stories/3044.md
#3048 Fix timezone bug in scheduler → NorthwindApp/.planning/stories/3048.md
Kept existing:
#3040 Add dark mode toggle (already analyzed — kept existing spec)
Planning complete. Run /devsprint-execute to implement all stories.
> /devsprint-execute
╔══════════════════════════════════════════════════════╗
║ Pre-flight Status Check ║
╚══════════════════════════════════════════════════════╝
Already completed:
✓ #3040 — Add dark mode toggle
Executed: 2026-03-08 | PR: https://dev.azure.com/.../pullrequest/187
Will execute:
→ #3044 — Email notification preferences
State: New | Tasks: 0/4 done | Repo: NorthwindApp
→ #3048 — Fix timezone bug in scheduler
State: New | Tasks: 0/2 done | Repo: NorthwindApp
Summary: 2 to execute, 1 already done
━━━ [1/2] Story #3044 — Email notification preferences ━━━
Created branch feature/3044-email-notifications from develop
... implementing ...
Story #3044 resolved ✓
━━━ [2/2] Story #3048 — Fix timezone bug in scheduler ━━━
Created branch feature/3048-timezone-fix from develop
... implementing ...
Story #3048 resolved ✓
╔══════════════════════════════════════════╗
║ Execution Complete ║
╚══════════════════════════════════════════╝
Sprint: Sprint 7
Stories processed: 2 | Previously completed: 1
✓ #3044 — Email notification preferences
Tasks: 4/4 resolved | Tests: 31 passed, 0 failed
PR: https://dev.azure.com/.../pullrequest/188
✓ #3048 — Fix timezone bug in scheduler
Tasks: 2/2 resolved | Tests: 18 passed, 0 failed
PR: https://dev.azure.com/.../pullrequest/189
Previously completed:
✓ #3040 — Add dark mode toggle
Completed: 2026-03-08 | PR: https://dev.azure.com/.../pullrequest/187
All pull requests:
https://dev.azure.com/.../pullrequest/187
https://dev.azure.com/.../pullrequest/188
https://dev.azure.com/.../pullrequest/189
Review the PRs in Azure DevOps, leave comments, then run /devsprint-pr-fix for each:
> /devsprint-pr-fix 3044
... fixes 3 review comments, pushes ...
> /devsprint-pr-fix 3048
... fixes 1 review comment, pushes ...
/devsprint-setup → Connect to Azure DevOps (one-time)
/devsprint-sprint → See your sprint board
/devsprint-create <description> → Create stories & tasks from natural language
/devsprint-plan [story-id] → Analyze stories → generate specs
/devsprint-execute [story-id] → Implement, commit, push, create PR, resolve tasks
/devsprint-pr-fix <story-id> → Fix PR review comments automatically
The typical workflow:
/devsprint-sprint— see what's in the sprint/devsprint-create— add missing stories or tasks (optional)/devsprint-plan— analyze stories, auto-detect repos, generate implementation specs/devsprint-execute— implement everything, create PRs, resolve tasks automatically
- Claude Code CLI installed
- Node.js (no external dependencies — uses built-in modules only)
- An Azure DevOps Personal Access Token with these scopes:
vso.project— read project/iteration datavso.work— read work itemsvso.work_write— create/update work itemsvso.code— create pull requests
git clone https://github.com/kloppnr1/claude-devsprint-plugin.git
cd claude-devsprint-plugin
./install.shThis copies commands and the helper script to ~/.claude/. Restart Claude Code to pick up changes.
Run the setup command in any project directory:
/devsprint-setup
This prompts for:
- Organization URL — e.g.,
https://dev.azure.com/yourorgor justyourorg - Project name — your Azure DevOps project
- PAT — your Personal Access Token
- Team — auto-detected from the project's team list (you pick from options)
- Area path — auto-resolved from your team's settings (no manual input needed)
Team and area ensure that new work items land in the right place on the board. Credentials are stored in .planning/devsprint-config.json (PAT is base64-encoded). This file is project-local — add it to .gitignore.
/devsprint-test
Confirms your credentials work and have the required scopes.
A real-time web dashboard for monitoring your sprint. Shows story status, active agent runs, PR status, execution history, and git activity — all auto-updating.
Click any story to expand it and see PR status, reviewer votes, execution history, and available actions:
Completed stories show the full lifecycle — Plan, PR Fix, and Exec runs with progress bars:
The daily overview shows a chronological feed of all sprint activity:
node dashboard/server.cjs --cwd /path/to/your/projectOpens at http://localhost:3000. The --cwd must point to the project directory that contains .planning/.
- Story groups — stories are automatically classified into status groups:
- Active — agent is currently working on the story
- Awaiting verification — story is Resolved with a pending PR. Click Approve to close it after staging verification
- Partial — execution started but didn't complete
- Planned — analyzed and ready for execution. Click Execute to start
- Not planned — in sprint but not yet analyzed. Click Analyze to plan it
- Blocked — title contains "BLOKERET" or execution was skipped
- Done — Closed/Done in Azure DevOps
- Live agent status — see what the agent is doing right now (which step, which story)
- Expandable run history — click completed runs to see step-by-step log with timestamps
- PR status — badges show active/merged/rejected, with a "Fix PR" button for active PRs
- One-click actions — Plan, Execute, Re-analyze, Fix PR, and Approve directly from the dashboard
- Auto-refresh — polls every 5 seconds for agent status, 60 seconds for full data refresh
The dashboard reads from files in .planning/:
devsprint-task-map.json— which stories are planned and wheredevsprint-execution-log.json— execution results per storydevsprint-agent-status.json— real-time agent progress
It also queries Azure DevOps for live sprint data and PR status.
Interactive setup wizard for Azure DevOps credentials. On re-run, shows current values (PAT masked) and lets you update individual fields.
Tests the stored credentials against the Azure DevOps API. Verifies both vso.project and vso.work scopes. Shows a clear success/failure message with suggested fixes.
Fetches and displays the current sprint backlog with ANSI-colored state indicators. Shows:
- Sprint name and dates
- User stories with description and acceptance criteria
- Tasks grouped under their parent story
- State (blue=Active, green=Resolved, etc.), assigned user, and other metadata
Defaults to showing only your assigned items (--me). Use --all to see the entire sprint.
Create stories and tasks from a natural language description. Parses your intent, creates work items via the Azure DevOps API — all assigned to the current sprint.
Examples:
/devsprint-create Add notification preferences page— creates a single User Story/devsprint-create Add CSV export: 1) Backend endpoint 2) Frontend button 3) Tests— creates a story with 3 tasks/devsprint-create Add tasks to #1205: write tests, update docs— creates tasks under an existing story
The main analysis pipeline. Run without arguments to plan all assigned stories, or pass a story ID to plan a single story (e.g., /devsprint-plan 1205). Already resolved stories are automatically skipped. Previously analyzed stories are skipped unless --reanalyze is passed.
What it does, step by step:
- Fetch stories — pulls your assigned stories from the current sprint via Azure DevOps API
- Auto-detect repos — checks task map history, sibling directories, and project repos. Only asks the user when auto-detection fails (first time for a new repo)
- Deep repo analysis — for each story, searches the target repo for files matching story keywords, reads relevant code, traces call chains, and checks for existing feature branches
- Show analysis — presents its understanding of each story (summary, work type, repo analysis, tasks). Non-research stories continue automatically; research stories get an interactive dialogue
- Update Azure DevOps — replaces the story description and acceptance criteria with the verified analysis (see warning below)
- Generate story spec — writes
STORY.mdto{repoPath}/.planning/stories/{storyId}.mdwith: goal, background, testable acceptance criteria, key files with paths, architecture/code flow, implementation notes, contacts, open questions, and out-of-scope items - Self-review — checks the generated spec against a quality checklist (specific goal, real file paths, traced code flow, no vague placeholders, blockers captured) and fixes issues before presenting
- Spec review — shows the full spec and asks "Changes?" — type corrections or "ok" to continue
- Post to Azure DevOps — adds a summary of the approved spec as a comment on the story
- Task map — writes/merges
.planning/devsprint-task-map.jsonmapping story IDs → repos → task IDs for status tracking during execution
Warning: Azure DevOps fields are overwritten. By default,
/devsprint-planreplaces the Description and Acceptance Criteria fields on each story with its verified analysis. Azure DevOps keeps revision history, so nothing is lost — but the original text is no longer visible without checking the history. If you want to keep the original fields untouched, add--no-devops-updateto skip all Azure DevOps writes (description, acceptance criteria, and comments). The local STORY.md spec is still generated.
Research mode: Stories tagged with research in Azure DevOps get a deeper treatment — broader codebase exploration, a multi-round dialogue where you discuss findings and possible approaches together, and a STORY.md that includes a "Research Findings" section with problem analysis, approaches considered, and the agreed approach.
Execute story plans. Two modes depending on arguments:
/devsprint-execute 1205— single story. Creates a feature branch, implements the story spec, resolves tasks, creates a PR. Mostly autonomous — only asks for input on items explicitly marked as blocking in the spec./devsprint-execute— all stories, fully autonomous. Loops through every story in the task map without user interaction. Errors on one story don't block the next. Outputs a full summary with all PR links at the end.
PRs are created via the Azure DevOps REST API and automatically linked to the story via workItemRefs.
Fix PR review comments. Takes a story ID, finds the matching PR automatically (by title prefix #{storyId}), fetches unresolved review comments, checks out the PR branch, fixes all issues, runs tests, and pushes.
/devsprint-pr-fix 1205
The devsprint-tools.cjs script handles all Azure DevOps API communication. It can be used standalone:
# Credentials
node devsprint-tools.cjs save-config --org <url> --project <name> --pat <token> [--team <team>] [--area <area>] --cwd <path>
node devsprint-tools.cjs load-config --cwd <path>
node devsprint-tools.cjs test --cwd <path>
# Teams & areas
node devsprint-tools.cjs list-teams --cwd <path>
node devsprint-tools.cjs get-team-area --team "<team name>" --cwd <path>
# Sprint
node devsprint-tools.cjs get-sprint --cwd <path>
node devsprint-tools.cjs get-sprint-items [--me] --cwd <path>
node devsprint-tools.cjs show-sprint [--me] --cwd <path>
# Work items
node devsprint-tools.cjs create-work-item --type <type> --title "<title>" [--description "<html>"] [--parent <id>] [--sprint] [--assigned-to "<name>"] [--area "<path>"] [--tags "<tags>"] --cwd <path>
node devsprint-tools.cjs update-state --id <workItemId> --state <state> --cwd <path>
node devsprint-tools.cjs update-description --id <workItemId> --description "<text>" --cwd <path>
node devsprint-tools.cjs update-acceptance-criteria --id <workItemId> --criteria "<html>" --cwd <path>
node devsprint-tools.cjs get-child-states --id <storyId> --cwd <path>
# Comments
node devsprint-tools.cjs add-comment --id <workItemId> --text "<html>" --cwd <path>
node devsprint-tools.cjs delete-comment --id <workItemId> --comment-id <commentId> --cwd <path>
# Git
node devsprint-tools.cjs create-branch --repo <path> --story-id <id> --title <title> [--base <branch>]
node devsprint-tools.cjs create-pr --repo <path> --branch <name> --base <branch> --title <title> --body <body> --story-id <id> --cwd <path>
# Pull requests
node devsprint-tools.cjs find-pr --story-id <id> --cwd <path> # Find PR by story ID (searches all repos)
node devsprint-tools.cjs get-pr --pr-id <id> [--repo-name <name>] --cwd <path> # Fetch PR details
node devsprint-tools.cjs get-pr-threads --pr-id <id> [--repo-name <name>] [--active-only] --cwd <path>
# Repositories
node devsprint-tools.cjs list-repos [--top <N>] --cwd <path>
node devsprint-tools.cjs get-branch-links --id <workItemId> --cwd <path>All commands output JSON to stdout and use exit code 0/1 for success/failure.
claude-devsprint-plugin/
├── install.sh # One-command installer
├── bin/
│ └── devsprint-tools.cjs # Node.js helper — all Azure DevOps API calls
├── commands/
│ ├── devsprint-setup.md # /devsprint-setup — credential configuration
│ ├── devsprint-test.md # /devsprint-test — connection verification
│ ├── devsprint-sprint.md # /devsprint-sprint — sprint backlog display
│ ├── devsprint-create.md # /devsprint-create — create stories & tasks
│ ├── devsprint-plan.md # /devsprint-plan — story analysis & spec generation
│ ├── devsprint-execute.md # /devsprint-execute — story execution & PR creation
│ └── devsprint-pr-fix.md # /devsprint-pr-fix — fix PR review comments
├── CONTRIBUTING.md
└── README.md
- PAT is stored base64-encoded in
.planning/devsprint-config.json— this is light obfuscation, not encryption - Always add
devsprint-config.jsonto.gitignore— the setup command checks for this - The PAT never leaves your local machine except in API calls to your Azure DevOps instance
- No external dependencies — the helper script uses only Node.js built-in modules (
https,fs,path)
MIT