Kaizen Agentic

AI agency framework: 20 specialized agents that arrive in your project informed, learn from experience, and improve over time.

kaizen-agentic provides two things: a library of agent instruction sets you deploy into projects, and an agency framework that gives those agents persistent memory and coordination. Agents accumulate project-scoped knowledge across sessions. A Coach meta-agent synthesises patterns across the entire fleet and briefs incoming agents on what to know first.

This project embraces the Japanese concept of "kaizen" (continuous improvement) applied to AI agent development. Every agent becomes part of an optimization loop where performance is measured, patterns are analyzed, and knowledge is carried forward.

Quick Start

Install the Package

From Source (Development):

git clone https://forgejo.coulomb.social/coulomb/kaizen-agentic.git
cd kaizen-agentic
make setup-complete
make agents-install-cli
source .venv/bin/activate  # Required for each session

Global Installation (Available from any directory):

git clone https://forgejo.coulomb.social/coulomb/kaizen-agentic.git
cd kaizen-agentic
make setup-complete
python3 -m build && make install-global
# No virtual environment activation needed

Local Package Testing:

git clone https://forgejo.coulomb.social/coulomb/kaizen-agentic.git
cd kaizen-agentic
make setup-complete
python3 -m build && make install-local
source .venv/bin/activate  # Required for each session

From Forgejo PyPI (current release: v1.4.0):

pip install kaizen-agentic \
  --extra-index-url https://forgejo.coulomb.social/api/packages/coulomb/pypi/simple/

# or global CLI via pipx
pipx install kaizen-agentic \
  --pip-args="--extra-index-url https://forgejo.coulomb.social/api/packages/coulomb/pypi/simple/"

See docs/PACKAGE_RELEASE.md for release and CI details.

Your First Project (New Users)

👋 New to Kaizen Agentic? Follow our Hello World Tutorial for a complete step-by-step guide.

Create a Project (Experienced Users)

# Create a new project with AI agents
kaizen-agentic init my-project --template python-web
cd my-project

# Set up development environment
make setup-complete

# Start coding with agent assistance!
make help  # See all available commands

Add Agents to Existing Project

# Navigate to your project
cd your-existing-project

# Install relevant agents
kaizen-agentic install keepaTodofile keepaChangelog tdd-workflow

# Check what was installed
kaizen-agentic status

Agency Framework

Agents deployed into a project can accumulate project-scoped memory — a structured file written at session close and read at session start. A Coach meta-agent reads across all agent memories and produces targeted orientation briefs for incoming agents.

# Scaffold memory for an agent
kaizen-agentic memory init sys-medic

# Brief an incoming agent using all existing project memories
kaizen-agentic memory brief tdd-workflow

# Review an agent's accumulated knowledge
kaizen-agentic memory show project-assistant

See docs/agency-framework.md for the full model.

Orientation

Read in this order for strategic context:

1. INTENT.md — purpose, boundaries, design principles
2. wiki/KaizenAgenticMission.md — product narrative
3. wiki/AboutKaizenAgents.md — agent concepts and metrics pilot
4. wiki/EcosystemIntegration.md — ecosystem composition
5. SCOPE.md — repository boundaries and current state
6. history/ — persisted assessments and gap analyses

Released v1.4.0 — see CHANGELOG.md. Repository state and completed delivery are summarized in SCOPE.md and WORK-RECORDS.md.

Feedback: kaizen-agentic feedback · docs/FEEDBACK.md

Features

• 20 Specialized Agents: Planning, testing, code quality, infrastructure, release, and meta-agent craft
• Agency Framework: Project-scoped agent memory + Coach meta-agent for cross-agent synthesis
• Metrics and Improvement Loop: Session evidence, optimizer recommendations, and optional artifact publication
• Roles and Engagements: Versioned role packages, client-bound engagement records, lifecycle checklists, and custody controls
• Scheduled Preparation: Repo-local schedules and offline orientation bundles for activity-core/Glas execution
• CLI Tool: Agent, memory, metrics, protocol, engagement, and schedule commands (kaizen-agentic)
• Project Templates: Pre-configured setups for different project types
• Runtime-neutral Contracts: Instruction and preparation surfaces usable by governed coding-agent harnesses
• Comprehensive Testing: Full test coverage with multiple testing strategies

Available Agents

Project Management

• keepaTodofile: Manages TODO.md files following Keep a Todofile format
• keepaChangelog: Maintains CHANGELOG.md files following Keep a Changelog format
• keepaContributingfile: Creates and updates CONTRIBUTING.md files
• project-assistant: General project planning and coordination
• priority-evaluation: Evaluates and orders competing work
• requirements-engineering: Requirements analysis and documentation
• scope-analyst: Maintains explicit product and repository boundaries
• releaseManager: Coordinates release readiness and delivery

Development Process

• tdd-workflow: Test-driven development workflow guidance
• test-maintenance: Test suite maintenance and optimization
• testing-efficiency: Improves test feedback speed and signal quality

Code Quality

• code-refactoring: Code improvement and refactoring guidance
• optimization: Agent definition optimization and improvement
• datamodel-optimization: Data model design and optimization
• tooling-optimization: Improves effective use of repository tooling

Infrastructure

• setupRepository: Repository initialization and standards compliance
• claude-documentation: Claude Code configuration and documentation
• sys-medic: Infrastructure health monitoring and diagnostics

Meta

• coach: Coaching meta-agent — reads all project agent memories, synthesises cross-agent briefs, and orients incoming agents
• wisdom-encouragement: Encourages reflective, evidence-backed improvement

Automated execution boundary

Kaizen Agentic defines agent craft, validates repo-local schedules, and prepares offline execution bundles. In the current unattended path, activity-core creates durable, claimable ops_run work. A Kaizen blueprint or agent instance hands a versioned harness_profile_ref to glas-harness, which resolves the concrete rein, model route, sandbox, tool policy, and limits and returns one evidence envelope. State Hub supplies roster and coordination evidence.

Manual execution remains supported. Kaizen Agentic itself does not own cron, the durable work queue, inference, credentials, or runtime authorization. See Integration Patterns and ADR-005.

View complete agent list

Project Templates

# Available templates
kaizen-agentic templates

# python-basic: Basic Python project setup
# python-web: Web application development
# python-cli: Command-line tool development
# python-data: Data science and analysis
# comprehensive: All available agents

Known Issues

Click Library Workaround

The CLI currently implements a workaround for spurious error messages in the Click library. This affects the install command but is transparent to users. See CLICK_WORKAROUND.md for technical details and removal timeline.

User Impact: None - the workaround provides clean CLI output Status: Monitoring Click library updates for resolution