Skip to main content

SpecWeave

The AI Development Framework That Can Run for Hours Autonomously.

Ship features while you sleep. /sw:auto executes tasks, runs tests, fixes failures, and syncs to GitHub/JIRA — completely hands-off. Mobile apps, microservices, multi-repo architectures — one framework handles it all.

NPM Version License: MIT Test & Validate Discord YouTube

# Requires Node.js 20.12.0+ (we recommend Node.js 22 LTS)
npm install -g specweave

New in v1.0.100: Auto mode now shows live labels in your terminal — see exactly what SpecWeave is doing: [Planning], [Implementing], [Testing], [Fixing]. Hours of autonomous execution with full visibility.


The Problem With AI Coding Tools

Every AI coding tool promises productivity. But after the chat ends:

What HappensThe Cost
Specs disappear into chat historyNo traceability for audits
Architecture decisions forgottenRepeated debates, inconsistent code
Tests? "We'll add them later"Regressions, production bugs
JIRA/GitHub manually updatedStale tickets, wrong priorities
Onboarding: "Ask John, he knows"Knowledge silos, bus factor

The real problem: AI generates code, but code without context is technical debt waiting to happen.


SpecWeave: AI That Works While You Sleep

/sw:increment "Add OAuth authentication"  # Creates spec + plan + tasks
/sw:auto # Autonomous execution for HOURS
# ↑ Watch live: [Planning] → [Implementing] → [Testing] → [Fixing] → [Done]

What happens during /sw:auto:

[08:23:41] [Planning]      Analyzing T-003: Implement refresh token rotation
[08:24:12] [Implementing] Writing src/auth/token-manager.ts (127 lines)
[08:25:33] [Testing] Running npm test -- token-manager.test.ts
[08:25:47] [Fixing] Test failed: Expected 401, got 403. Adjusting...
[08:26:15] [Testing] Re-running tests... PASSED
[08:26:18] [Syncing] Updating tasks.md, pushing to GitHub
[08:26:22] [Done] T-003 complete. Moving to T-004...

Or step-by-step control:

/sw:do                                     # Execute one task at a time
/sw:done 0001 # Validates: tests pass, docs updated

Every feature creates three permanent files:

.specweave/increments/0001-oauth/
├── spec.md <- WHAT: User stories, acceptance criteria
├── plan.md <- HOW: Architecture decisions, tech choices
└── tasks.md <- DO: Implementation tasks with embedded tests

After 6 months: Search "OAuth" -> find exact decisions, who approved, why it was built that way.


Why SpecWeave vs BMAD, SpecKit, Cursor Rules?

CapabilityBMADSpecKitCursor RulesSpecWeave
MaturityAlpha (v6)StableBuilt-inProduction (v1.0.100)
Autonomous WorkManual steps requiredOne-shot generationPer-request onlyHours unattended
Live StatusNoneNoneNoneTerminal labels
Multi-RepoSingle repoSingle repoSingle repoCoordinate N repos
Mobile AppsNo agentsNo agentsNo agentsReact Native/Expo
External SyncNoneNoneNoneGitHub/JIRA/ADO
BrownfieldLimitedGreenfield onlyAny10-year legacy? Fine.
Quality GatesNot built-inNoneNone3-gate validation
Living DocsManualSnapshotNoneAuto-update
Self-LearningNoneNoneNoneReflects on mistakes
DogfoodingUnknownUnknownN/A140+ self-built features

The math: SpecKit output = ONE SpecWeave increment. SpecWeave = N increments + lifecycle + sync + hooks + hours of autonomous execution.


Proof: Built With SpecWeave

This framework builds itself. Every feature, bug fix, and release — all spec-driven.

Deploy Frequency Lead Time Change Failure Rate MTTR

MetricValue
Features shipped140+ with full specs
Living docsAuto-update on every task
DORA metricsLive tracking
Test coverage60%+ enforced

Browse our increments ->


Quick Start

Prerequisites

Node.js 20.12.0 or higher is required (we recommend Node.js 22 LTS).

# Check your version
node --version

# If below v20.12.0, upgrade using nvm:
nvm install 22 && nvm use 22 && nvm alias default 22

Getting SyntaxError: Unexpected token 'with'? Your Node.js is too old. See upgrade instructions.

Installation

npm install -g specweave
cd your-project
specweave init .

Then in Claude Code:

/sw:increment "Add dark mode"  # AI creates spec + plan + tasks
/sw:auto # Ship while you sleep (hours of autonomous work)

Or step-by-step:

/sw:do                         # Execute tasks one at a time
/sw:done 0001 # Quality-validated completion

Troubleshooting - if commands/skills stop working:

specweave refresh-marketplace   # Reinstall all plugins from GitHub
specweave update-instructions # Regenerate CLAUDE.md

Full Quickstart Guide ->


Key Features

Autonomous Execution (v2.9)

Run for hours without intervention. See exactly what's happening:

/sw:auto                  # Start autonomous mode
/sw:auto-status # Check progress anytime
/sw:cancel-auto # Emergency stop (rarely needed)

What you see in terminal:

  • [Planning] - Analyzing task requirements
  • [Implementing] - Writing code
  • [Testing] - Running test suites
  • [Fixing] - Auto-correcting failures
  • [Syncing] - Updating docs and external tools
  • [Done] - Task complete, moving to next

External Tool Integration

PlatformCapabilities
GitHub IssuesCreate, update, close, progress sync
JIRAEpic/Story hierarchy, bidirectional sync
Azure DevOpsWork items, area paths, status sync
/sw:sync-progress  # Push updates to all connected tools

Works With ANY AI Tool

SpecWeave is structure, not AI magic. spec.md and tasks.md are just markdown.

  • Sarah uses Claude. Mike uses GPT-4. Alex uses Copilot.
  • Everyone works on the same specs, same tasks, same acceptance criteria.
  • Claude Code gets the best experience (slash commands, hooks, skills). But any AI can participate.

Living Documentation

Documentation updates on lifecycle events — not on every edit, but when it matters:

EventWhat Updates
increment.doneMarked complete, living docs synced
user-story.completedStatus refreshed, external tools updated
increment.reopenedRestored from archive

Quality Gates

Three gates before any increment closes:

  1. Tasks: All marked complete
  2. Tests: 60%+ coverage (configurable)
  3. Documentation: Living docs updated

Token Efficiency

70%+ context reduction:

  • Progressive plugin loading
  • Skills auto-activate on keywords
  • Context optimizer removes irrelevant specs

Commands

CommandPurpose
/sw:increment "feature"Create spec + plan + tasks
/sw:autoShip while you sleep - hours of autonomous work
/sw:auto-statusCheck autonomous session progress
/sw:doExecute one task at a time
/sw:done 0001Close with quality validation
/sw:sync-progressPush to GitHub/JIRA/ADO
/sw:cancel-autoStop autonomous session

53 total commands ->


Works On Everything

ScenarioWhat You Do
10-year legacy codebasespecweave init . -> brownfield analysis detects doc gaps
Weekend MVPspecweave init . -> /sw:increment "Build auth"
50-team enterprisespecweave init . -> /sw:enable-multiproject -> maps to JIRA/ADO

Requirements

  • Node.js 20.12.0+ (we recommend Node.js 22 LTS) — upgrade guide
  • Any AI coding tool: Claude Code, Cursor, Windsurf, Cline, Aider, etc.
  • Git repository

Why Node.js 20.12.0+? SpecWeave uses modern JavaScript features (ES2022 Import Attributes) that require Node.js 20.12.0 or higher. If you see SyntaxError: Unexpected token 'with', you need to upgrade Node.js.


Community

Documentation | Discord | YouTube | GitHub Issues


License

MIT — github.com/anton-abyzov/specweave