Skip to main content

SpecWeave

The spec-driven framework for AI coding agents. First-class support for Claude Code — compatible with any LLM-powered coding tool.

Coordinate parallel AI agents. Prevent task overlap. Quality gates. Ship features while you sleep.

NPM Version License: MIT Test & Validate Discord

npm install -g specweave   # Requires Node.js 20.12.0+

Quick Demo

/sw:increment "User authentication"   # AI creates spec + plan + tasks
/sw:auto # Autonomous execution for hours
/sw:grill 0001 # Senior-level code review
/sw:done 0001 # Validate and complete

What happens: Claude creates a full specification, executes tasks autonomously, runs tests, fixes failures, and syncs progress to GitHub/JIRA. You review finished work.


Why SpecWeave?

AI coding agents are powerful individually. Run three of them on the same codebase and you get conflicts, duplicated work, and zero traceability.

SpecWeave solves this with file-based coordination:

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

Each increment is a self-contained scope. Each agent knows exactly what's taken and what's available. No overlap. No conflicts. Six months later, search "OAuth" and find exactly what was decided, who approved it, and why.

What Makes It Different

  • Parallel agent coordination — multiple Claude Code sessions, OpenClaw instances, or remote agents work on different increments without stepping on each other.
  • Autonomous execution — runs for hours, not minutes. Write, test, fix, repeat.
  • Persistent memory — AI learns from corrections. Fix once, remembered permanently.
  • Quality gates — Code Grill reviews code like a demanding senior engineer before every release.
  • Living documentation — specs, ADRs, and runbooks sync automatically after every task.
  • 100+ specialized skills — PM, Architect, QA, Security, DevOps, Frontend, Mobile, ML, and more collaborate on deliverables.
  • External sync — GitHub Issues, JIRA, Azure DevOps — bidirectional, real-time.
  • Self-improving — captures what works and what doesn't. Gets smarter over time.
  • LSP integration — semantic code intelligence for TypeScript, Python, Go, Rust, Java, C#.

Install

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

Then in Claude Code:

/sw:increment "Add dark mode"   # Describe your feature
/sw:auto # Ship while you sleep

Node.js 20.12.0+ required (22 LTS recommended). Getting SyntaxError? Upgrade instructions.


Parallel Development

Run multiple AI agents on the same repository — locally, in the cloud, or with OpenClaw. SpecWeave's increment files are the coordination layer:

Agent 1 (local Claude Code)    Agent 2 (cloud)           Agent 3 (OpenClaw)
working on: 0001-auth working on: 0002-payments working on: 0003-notifications
│ │ │
└──────────────────────────────┼───────────────────────────┘

.specweave/increments/
├── 0001-auth/tasks.md ← Agent 1's scope
├── 0002-payments/tasks.md ← Agent 2's scope
└── 0003-notifications/tasks.md ← Agent 3's scope

How it works:

  1. Create increments for each feature: /sw:increment "auth", /sw:increment "payments"
  2. Each agent picks an increment and runs /sw:auto — tasks.md tracks exactly what's done
  3. Agents work in isolated scopes — different files, different specs, different tests
  4. Quality gates (/sw:grill) ensure consistent standards regardless of which agent built it
  5. Progress syncs to GitHub/JIRA so you see everything in one place

Why this matters: OpenClaw and Claude Code sessions are stateless by default. SpecWeave's spec/plan/tasks files persist across sessions and agents — your coordination layer survives restarts, crashes, and context window limits.

Agent team commands:

CommandPurpose
/sw:team-orchestrate "feature"Split feature across parallel agents
/sw:team-statusMonitor all agent progress
/sw:team-mergeMerge completed work in dependency order

Full agent teams guide


Core Commands

CommandPurpose
/sw:increment "feature"Create spec + plan + tasks
/sw:autoAutonomous execution
/sw:doExecute one task at a time
/sw:grillCode review before close
/sw:doneClose with quality validation
/sw:sync-progressPush to GitHub / JIRA / ADO
/sw:nextAuto-close + suggest next

Full command reference


Integrations

PlatformWhat Syncs
GitHubIssues, PRs, milestones — bidirectional
JIRAEpics, stories, status
Azure DevOpsWork items, area paths

When you close an increment, external tools update automatically.


How It Compares

CapabilitySpecWeaveBMAD MethodGitHub SpecKit
Parallel agent coordinationIncrement-scoped isolationNoNo
Autonomous executionHours of unattended /sw:autoNoNo
Quality gatesCode Grill before every releaseNoNo
Living documentationAuto-updated after every taskManualManual
Self-improving AILearns from correctionsNoNo
External syncGitHub / JIRA / ADO bidirectionalNoNo
Specialized skills100+ (PM, QA, DevOps, ML...)21 agentsNone
Works with OpenClawYesPartialPartial
Spec/plan/tasks workflowYesYesYes
Agent-agnosticClaude Code + OpenClaw + Copilot + CodexMulti-IDEMulti-IDE

Built With SpecWeave

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

Browse increments — see how SpecWeave develops SpecWeave.


Documentation

spec-weave.com — guides, examples, and full reference.

Community

Discord · YouTube · GitHub Issues

License

MIT — github.com/anton-abyzov/specweave