1.8.2 (2026-02-17)
- harden session-start.js parallel detection (5c7f05b)
- Windows compatibility and correct hook configuration syntax (1e8b22c)
1.8.1 (2026-02-12)
- docs: Fix broken search with Pagefind plugin (0600ef0)
- hooks: add explicit break for no-fallthrough lint (55f000c)
- hooks: fix exit code in default case - changed to process.exit(0) (ada6acb)
1.8.0 (2026-02-08)
- add pattern learning feature for agents to compound knowledge across sessions (ac5946b)
1.7.0 (2026-02-04)
- canvas: add missing viewportSize destructuring in EditorCanvas (7a01a8e)
- resolve init warnings for research skill and native binary detection (0e413ab)
- sidebar: default to sidebar mode to prevent flash of wrong content (d5e2610)
- tools: correct feature flags based on official docs (1818af3)
- tools: Cursor has skills (https://cursor.com/docs/context/skills) (d4b5864)
- use Anchor icon instead of Hook (lucide-react compat) (4187028)
- canvas: add viewport size toggles (desktop, tablet, mobile) (4964e23)
- dev: add automatic hot reload for extension development (b70e34b)
- icons: use dashboard-icons CDN for tool and MCP icons (a26f9a6)
- tools: add comprehensive AI CLI tool detection with Aider, Cline, Kilo (38b6f27)
1.6.0 (2026-01-29)
- add AI provider abstraction layer (b583d4e)
- add MCP canvas resources and monorepo structure (5c593f5)
- add ShadCN component registry (722c92a)
- add VS Code extension infrastructure (4a5b33f)
- convert studio package to VS Code extension with webview (4fa4c1d)
- fix all remaining TypeScript errors and achieve successful build (778b030)
- fix TypeScript errors in studio package (fa2105b)
- merge VS Code extension infra (3676a43)
- remove chat panel and update layout system (ffb499b)
- studio: merge studio into monorepo and add Integration Hub (06b6d2f)
- studio: update extension with theme sync, hot reload, and Integration Hub (8845863)
1.5.1 (2026-01-28)
- correct hook paths from .claude/hooks to bin/hooks (c616f2a)
1.5.0 (2026-01-28)
- add language specifiers to MCP code blocks in agent files (a6a9edd)
- add missing hooks to template directory (f304e06)
- include schema.sql in MCP server build (28a4433)
- support both snake_case and camelCase in architecture.json health check (1b6cd4d)
- add circuit breaker pattern and UX improvements (5789487)
- Add integration and unit tests for MCP server and related components (57b71eb)
- detect NEW libraries (not just version changes) for skill generation (fcd8f86)
- integrate MCP server with agentful agents (d9d7943)
- make architect drift detector smarter and less noisy (9dcd18e)
- make architect drift detector truly useful with smart filtering (80d36b7)
1.4.1 (2026-01-28)
- prevent corruption of native Claude Code binary on macOS (29d1573)
1.4.0 (2026-01-25)
- improve responsiveness of comparison grid for smaller screens (050e955)
- update module description for clarity and accuracy (531a837)
- update references in state validator documentation for clarity (c42239f)
- add documentation for /agentful-init command and update navigation links (ef1c6c5)
- enhance onboarding and command structure in documentation; add hooks for file creation and health checks (ee7316d)
- enhance responsiveness of comparison and decision grids in CSS (03cd6b5)
1.3.0 (2026-01-25)
- docs: add DiagramModal component for interactive architecture and workflow visualizations (ed9c84c)
- docs: enhance architecture documentation with interactive DiagramModal components (6a96ac0)
- docs: update architecture and background agents documentation with mermaid diagrams for better visualization (b44fb8f)
- mcp: complete distributed MCP architecture with horizontal scaling (2e36b5b), closes #distributed-mcp #horizontal-scaling #oauth-authentication
- mcp: implement MCP server with stdio transport and fix hook paths (865132d), closes #distributed-mcp-foundation
- update version to 1.2.2 and enhance agent generation workflow (f523a86)
1.2.1 (2026-01-23)
1.2.0 (2026-01-23)
- Add documentation for local models and self-hosted agent server (657d4fc)
1.1.0 (2026-01-23)
- Update README and documentation for agentful toolkit (247fc4e)
1.0.2 (2026-01-23)
- daemon: add log file redirection and health check wait (ec7097d)
- server: allow network connections with --auth=none (35bac75)
1.0.1 (2026-01-23)
- server: critical usability improvements for remote execution (bb77a78)
1.0.0 (2026-01-23)
- comprehensive audit and fix of all documentation inaccuracies (ffea5b8)
- State file schemas updated to match documented behavior. Existing .agentful/ directories will need reinitialization.
0.5.1 (2026-01-23)
- trigger npm publish on push to main, let semantic-release manage versioning (cc0290b)
- NEW:
--daemon/-dflag foragentful servecommand- Runs server in background as a detached process
- Automatically creates PID file at
.agentful/server.pid - Returns immediately after starting daemon
- NEW:
agentful serve --stopto stop daemon- Gracefully shuts down daemon using SIGTERM
- Falls back to SIGKILL if needed
- Automatically cleans up PID file
- NEW:
agentful serve --statusto check daemon status- Shows PID, uptime, and memory usage (when available)
- Detects and reports stale PID files
- Provides helpful commands for management
- IMPROVED: Better process management
- Prevents starting multiple daemons on same port
- Detects stale PID files and cleans them up
- Handles permission errors gracefully
# Start server in background
agentful serve --daemon
# Check if daemon is running
agentful serve --status
# Stop daemon
agentful serve --stop
# Start daemon with custom options
agentful serve --daemon --auth=hmac --secret=xxx --port=3737- BREAKING: Removed confusing tech-stack-specific presets
- β Removed:
fullstack-typescript,fullstack-javascript,fullstack-python,enterprise - β
Kept:
default(all components) andminimal(orchestrator + backend)
- β Removed:
- NEW: Default behavior installs all components (no flags needed)
npx @itz4blitz/agentful initβ installs everything (recommended)- Tech stack is auto-detected on first run, irrelevant to installation
- IMPROVED: Clearer messaging and philosophy
- "One product: agentful" - not multiple flavors
- "Default to power" - give users everything, let them remove what they don't need
- Tech-agnostic approach - works with any stack
- Simplified README.md installation section
- Updated help text and examples in CLI
- Updated
agentful presetscommand output - Added philosophy section explaining the simplification
# Install agentful (all components - recommended)
npx @itz4blitz/agentful init
# Minimal setup (for simple scripts/CLIs)
npx @itz4blitz/agentful init --preset=minimal
# Custom configuration
npx @itz4blitz/agentful init --agents=orchestrator,backend --skills=validation
# View installation options
npx @itz4blitz/agentful presets- NEW:
/agentful-updatecommand for safe configuration updates- Updates
.claude/directory without overwriting customizations - Preserves user modifications and project-specific configurations
- Intelligent merge strategy for template updates
- Updates
- NEW: Post-merge git hook for automatic
.claude/restoration- Auto-installs at
.git/hooks/post-merge - Automatically restores
.claude/directory after git pull/merge - Ensures configuration consistency across team updates
- Auto-installs at
- NEW: Template directory separation for development safety
- Templates stored in
template/.claude/directory - Working configurations in
.claude/directory - Prevents accidental template modifications during development
- Templates stored in
- NEW: Complete test coverage with 370 tests (100% passing)
- Unit tests for CLI and initialization logic
- Schema validation tests for JSON state files
- Agent markdown validation tests
- Command markdown validation tests
- NEW: Test coverage reporting (
npm run test:coverage)- 98.26% statement coverage on core library (
lib/init.js) - 100% function coverage on core library
- Multiple reporters: text, json, html, lcov
- 98.26% statement coverage on core library (
- NEW: Vitest configuration with comprehensive settings
- NEW: Testing strategy documentation (
TESTING_STRATEGY.md)- Explains what's tested vs. what's not testable
- Provides roadmap for future testing improvements
- Documents hybrid testing approach for AI-driven systems
- NEW: Pre/post hooks for agent invocations and feature completion
pre-agent.js- Validates preconditions before agent executionpost-agent.js- Tracks agent invocation metricspre-feature.js- Validates feature readinesspost-feature.js- Runs quality gates and creates commits
- NEW: Agent metrics tracking in
.agentful/agent-metrics.json - NEW: Comprehensive hooks documentation (
LIFECYCLE_HOOKS.md)
- NEW: Error handling in all agents
- Common error scenarios documented
- Recovery strategies defined
- Retry logic with exponential backoff
- Escalation procedures
- NEW: Error logging format standardization
- NEW: Agent timeout awareness and prevention strategies
- IMPROVED: ESLint configuration with proper ignores
- FIXED: All unused variables and imports removed
- FIXED: Indentation and quote style consistency
- IMPROVED:
.gitignoreupdated (coverage/, docs/dist/)
- IMPROVED: All agent markdown files now include:
- Error handling sections
- Scope definitions (Your Scope vs. NOT Your Scope)
- Comprehensive rules sections
- Retry strategies
- Escalation procedures
- IMPROVED: All command files use
//comments instead of#in code blocks - IMPROVED: Architecture documentation enhanced
- ENHANCED: Architect agent with error recovery
- ENHANCED: Backend agent with database connection handling
- ENHANCED: Frontend agent with component import error handling
- ENHANCED: Fixer agent with infinite loop detection
- ENHANCED: Orchestrator agent with robust state management
- ENHANCED: Product analyzer agent with proper Rules section
- ENHANCED: Reviewer agent with timeout awareness
- ENHANCED: Tester agent with TodoWrite tracking examples
- FIXED: CLI banner now includes "AGENTFUL" text for test validation
- FIXED: Multiple top-level headings in command markdown files
- FIXED: Code block fence pairing in agent markdown files
- FIXED: Placeholder text detection in validation tests
- FIXED: TODO comment detection in validation tests
- FIXED: Frontmatter parsing bug in validation tests
- FIXED: Unclosed markdown code block in architect.md
- FIXED: Missing language specifiers in code blocks
- β Zero ESLint Errors: Clean codebase
- β 370/370 Tests Passing: Full test coverage
- β 98.26% Core Coverage: Excellent code coverage
- β Zero Security Issues: No vulnerabilities
- β Production Ready: All quality gates passing
- β CI/CD Ready: GitHub Actions workflow configured
- Test Framework: Vitest with coverage via v8
- Test Count: 370 tests across 5 test files
- Coverage: 98.26% statements, 100% functions on core library
- Validation: Agent and command markdown structure validation
- Hooks: 4 lifecycle hooks for validation and metrics
- Error Handling: Comprehensive error recovery in all agents
No breaking changes. All additions are backward compatible.
- Test Files: 5 new test files added
- Agent Files: 8 agents enhanced with error handling
- Command Files: 8 commands improved
- Documentation: 3 new docs (TESTING_STRATEGY.md, LIFECYCLE_HOOKS.md, updated ARCHITECTURE.md)
- Configuration: ESLint config, .gitignore, vitest.config.js
- CI/CD: GitHub Actions workflow for automated testing
- Total Changes: 45 files, 9,791 insertions
0.4.0 (2026-01-20)
- remove PRODUCT.md, use hierarchical .claude/product/ only (44eea9a)
- PRODUCT.md at project root is no longer supported. All product specifications must now use .claude/product/ structure.
Changes:
- Remove PRODUCT.md template file
- Update init to create .claude/product/index.md with comprehensive template
- Update init to create .claude/product/README.md with structure docs
- Remove all PRODUCT.md detection and parsing logic
- Update all agents, commands, skills to use .claude/product/ only
- Update all documentation to reference hierarchical structure
Migration for existing projects:
- Move PRODUCT.md contents to .claude/product/index.md
- Or run /agentful-product to restructure
The hierarchical structure scales better for complex projects while still supporting simple flat structure via .claude/product/index.md.
0.3.0 (2026-01-20)
- cli: add /agentful-generate command for smart codebase analysis (9f9c53e)
0.2.1 (2026-01-20)
- address architectural issues across codebase (e20412b)
0.2.0 (2026-01-20)
- remove dead links in docs (e514b7b)
- remove dead links introduced by doc improvements (7e40894)
- typo and add /agentful-product to docs sidebar (bbfbd2c)
- upgrade @semantic-release/npm to v13 for OIDC support (31d1e9f)
- upgrade semantic-release to v25 for OIDC support (2eced0f)
- add /agentful-analyze command and trim bloated files (199b40b)
- major architecture overhaul - agents, skills, positioning, docs (51b504e)
0.2.3 (2026-01-18)
- revert docs to root path and fix landing page links (cb72858)
- update workflow to always publish and fix landing page (0760412)
0.2.2 (2026-01-18)
- publish npm package on every release attempt (460b487)
0.2.1 (2026-01-18)
- remove old HTML landing page in favor of vocs MD layout (4fb1b78)
0.2.0 (2026-01-18)
- restructure docs with landing page and add /agentful command (18f030c)
- add contents:write permission to Release workflow (debdd5a)
- add NPM_TOKEN to Release workflow (015be42)
- correct docs output path in Deploy workflow (bdcf8ce)
- streamline GitHub Actions for automated publishing (a389f08)
- support both NPM_TOKEN and OIDC for publishing (163e8e3)
- use GitHub Actions expression for NPM_TOKEN and cleanup docs (84a7305)
- use npm publish directly with OIDC instead of semantic-release (c71456f)
- add smart analysis, conversational interface, and dual product structure (1659d32)
All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- NEW:
/agentfulcommand for natural conversation with Agentful- Understands intent from natural language (no need to remember specific commands)
- Context-aware conversation history tracking
- Handles edge cases gracefully (ambiguous requests, context loss, mind changes)
- Integrates seamlessly with all existing commands
- NEW: Conversation skill with NLP capabilities
- Intent classification with confidence scoring (12+ intents)
- Reference resolution ("it", "that", "the feature")
- User preference learning (communication style, update frequency)
- Context loss recovery (>24h gaps)
- NEW: Conversation history state management
.agentful/conversation-history.jsonfor session tracking- Thread continuity across messages
- User preference storage
- NEW: Intelligent project analyzer
- Multi-phase analysis (quick scan, deep analysis, pattern mining)
- Language detection (15+ languages supported)
- Framework detection (40+ frameworks across all ecosystems)
- Database, testing, and styling detection
- Pattern mining from actual codebase
- NEW: Domain detection algorithm
- Multi-source signal aggregation (structure, API routes, database, modules)
- 18+ business domain patterns (auth, billing, content, e-commerce, etc.)
- Confidence scoring and threshold filtering
- Auto-discovers project organization
- ENHANCED:
agentful initnow includes smart analysis by default- Auto-detects tech stack and domains
- Shows confidence scores with color-coded output
- Option to generate specialized agents and domain structure
- Graceful fallback to basic init on errors
- NEW: Project-specific agent generation
- Core agents (backend, frontend, tester, reviewer, fixer) adapted to your stack
- Domain-specific agents (auth-agent, billing-agent, content-agent, etc.)
- Technology-specific agents (nextjs-agent, prisma-agent, django-agent, etc.)
- Real code examples extracted from your project
- Naming conventions and architectural patterns detected automatically
- NEW: Hierarchical domain structure generator
- Creates
.claude/product/domains/{domain}/index.md - Generates feature specifications with acceptance criteria
- Documents API endpoints, data models, and dependencies
- Supports migration from flat to hierarchical structure
- Creates
- β
17 documentation files updated to reflect dual product structure support
- Comprehensive comparison tables (flat vs hierarchical)
- Decision matrices for choosing the right structure
- Migration paths documented
- Real-world examples for both approaches
- Updated: project-structure.mdx (1,079 lines - comprehensive guide)
- Updated: writing-product-md.mdx (two complete templates)
- Updated: All getting-started guides (first-project, quick-start, configuration)
- Updated: Examples show hierarchical structure for production apps
- Updated: README.md with current product structure information
- Updated: All CLI help text and output messages
- NEW:
--smartflag (enabled by default) for smart init - NEW:
--no-smartflag to disable smart analysis (basic init) - NEW:
--deepflag for thorough analysis (future enhancement) - NEW:
--generate-agentsflag to auto-generate agents - NEW:
--generate-domainsflag to auto-generate domain structure - ENHANCED: Beautiful colored output with confidence bars
- ENHANCED: Interactive prompts for agent/domain generation
- ENHANCED: All help text explains both product structures
- Zero Breaking Changes: 100% backward compatible
- Enhanced Templates: Both PRODUCT.md and CLAUDE.md now explain both structures
- Better Error Handling: Graceful degradation throughout
- Performance: Analysis completes in <30 seconds for medium projects
- Type Safety: All new code follows ES Modules standards
- π Easier Onboarding: Start with simple PRODUCT.md, migrate to hierarchical when ready
- π Natural Interaction: Just talk to Agentful - no need to memorize commands
- π Project Awareness: Agentful understands your tech stack automatically
- π Better Organization: Hierarchical structure for larger projects
- π Auto-Detection: No configuration needed - Agentful adapts to your format
- New Files: 12 (lib/* analysis modules, .claude/commands/agentful.md, .claude/skills/conversation/*, etc.)
- Modified Files: 28 (docs, templates, CLI, config)
- Total Lines Added: ~4,000 lines of production-ready code
- Documentation: 17 files comprehensively updated
No migration required! Existing projects continue working with:
- Existing PRODUCT.md files
- All existing commands
- All existing state files
For new projects, Agentful now recommends:
- Small projects: PRODUCT.md at root (simple, flat structure)
- Large projects: .claude/product/domains/ (organized, hierarchical structure)
- Smart init has not been tested on diverse codebases yet
- Domain detection accuracy needs real-world validation
- No automated test suite yet (planned for v0.2.1)
- Some advanced features need more documentation
- Add automated test suite
- Performance optimization
- Enhanced error messages
- Domain customization guide
- Interactive init wizard
- Initial alpha release of agentful
- CLI tool for autonomous product development
- 7 specialized agents (architect, backend, frontend, orchestrator, reviewer, tester, fixer)
- 4 slash commands (agentful-start, agentful-status, agentful-decide, agentful-validate)
- 2 skills (product-tracking, validation)
- Project template with CLAUDE.md and PRODUCT.md
- Documentation site built with Vocs
- Support for both flat and hierarchical product structures
- Auto-detection of product specification format
- 24/7 autonomous development capabilities
- Comprehensive getting started guide
- Agent documentation
- Command reference
- Configuration guide
- Workflow examples
- Troubleshooting guide
npx @itz4blitz/agentful init- Initialize new agentful project- Automatic state tracking in
.agentful/ - Multi-format product spec support
- Hybrid product structure (flat and hierarchical)
- Auto-detection and format adaptation
- Completion tracking at domain, feature, and subtask levels
- Untested in real projects (alpha release)
- May have critical bugs
- APIs may change without notice
- Not recommended for production use
MIT License - See LICENSE file for details