Turn AI coding agents into disciplined engineers.
Português · Getting Started · Skills Catalog · Get Pro →
AI coding agents generate code fast. But without governance, you get:
- No tests, no review, no validation
- Deprecated APIs, broken imports, scope creep
- Code that works in demo but breaks in production
- Agents that ignore your architecture and conventions
AgentKit is a governance framework — skills, rules, agents, and pipeline scripts — that enforces engineering discipline on any AI coding agent. It doesn't generate code; it makes your agent generate good code.
# One-liner install
curl -sL https://raw.githubusercontent.com/Joaonic/agentkit/main/install.sh | bash| Asset | Free | Pro | Description |
|---|---|---|---|
| Skills | 14 | 67 | Executable step-by-step workflows |
| Rules | 10 | 51 | Always-on guardrails (auto-loaded by Cursor) |
| Agents | 4 | 11 | Specialized personas with domain expertise |
| Pipeline Scripts | — | 4 | Autonomous implement → review → merge loop |
| Design Patterns | — | 22 | All GoF patterns as executable skills |
| DB Workbenches | — | 3 | Postgres, Redis, MongoDB, Elasticsearch |
| Workflow Docs | — | 8 phases | Complete governance system documentation |
| Quality Gates | 2 | 5 | Mandatory checkpoints before completion |
┌─────────────┐ ┌──────────────┐ ┌──────────────┐ ┌──────────┐
│ PLAN │────▶│ IMPLEMENT │────▶│ REVIEW │────▶│ MERGE │
│ new-plan │ │ TDD-first │ │ code-review │ │ posttask │
│ qa-spec │ │ worktrees │ │ ux-review │ │ verify │
└─────────────┘ └──────────────┘ └──────────────┘ └──────────┘
Gate 1 Gate 2 Gate 3 Gates 4 & 5
5 Mandatory Quality Gates:
- Planning — No implementation without clear scope and acceptance criteria
- TDD — Behavior changes require failing-first tests
- Review — No completion without findings-first technical review
- Validation — No completion without command-level posttask evidence
- Documentation — Changes require doc synchronization
# Auto-detects GitHub/GitLab, installs free tier
curl -sL https://raw.githubusercontent.com/Joaonic/agentkit/main/install.sh | bash
# Explicit provider and tier
curl -sL https://raw.githubusercontent.com/Joaonic/agentkit/main/install.sh | bash -s -- --provider github --tier freegit clone https://github.com/Joaonic/agentkit.git
cd your-project
../agentkit/install.sh --provider github --tier free# Copy the package contents into your project root
cp -r agentkit-free-github/.cursor your-project/
cp -r agentkit-free-github/.github your-project/ # GitHub only
cp agentkit-free-github/AGENTS.md your-project/| Tool | How |
|---|---|
| Cursor | Native .cursor/rules/ + .cursor/skills/ |
| GitHub Copilot | .github/copilot-instructions.md + .github/skills/ |
| Windsurf | Via .windsurfrules (adapt rules manually) |
| Cline | Via .clinerules (adapt rules manually) |
| Any CLI agent | copilot-cli, aider, etc. — reads AGENTS.md |
| Provider | CLI | Commands |
|---|---|---|
| GitHub | gh |
gh issue create, gh pr create, gh pr merge |
| GitLab | glab |
glab issue create, glab mr create, glab mr merge |
Auto-detected based on your project (.github/ → GitHub, .gitlab-ci.yml → GitLab, remote URL).
brew install gh # macOS
gh auth login # Login via browser
gh auth status # Verifybrew install glab # macOS
glab auth login # Login via browser
glab auth status # Verify| Category | Skills |
|---|---|
| Planning | new-plan, implement-plan |
| Quality | tdd-workflow, posttask, posttask-workspace-lint, qa-issue-spec |
| Build & Test | run-tests, run-build, lint-fix |
| Review | review-open-pr |
| Meta | add-skill, bootstrap, bootstrap-governance, new-adr |
| Category | Skills |
|---|---|
| Planning | create-milestone, executing-plans-parallel, plan-to-issues, prioritize-roadmap, prioritize-github-roadmap, prioritize-gitlab-roadmap |
| Audit | code-audit, code-audit-architecture-consistency, docs-audit, docs-audit-sync |
| Testing | e2e, ux-review |
| Backend (Java) | new-use-case, new-adapter, new-api-resource, db-migrate, gen-api |
| Backend (NestJS) | backend-hexagonal-nestjs, openapi-orval-nextauth |
| Frontend | new-feature, new-widget, new-ui-component, frontend-design |
| Infrastructure | postgres-workbench, redis-workbench, elasticsearch-workbench, mongo-workbench |
| AI/ML | ai-guardrails-and-redaction, ai-router-executor-pattern |
| Design Patterns | All 22 GoF: abstract-factory, adapter, bridge, builder, chain-of-responsibility, command, composite, decorator, facade, factory-method, flyweight, iterator, mediator, memento, observer, prototype, proxy, singleton, state, strategy, template-method, visitor |
| Platform | add-mcp, saas-mvp-bootstrap |
| Script | What it does |
|---|---|
pipeline.sh |
Full autonomous loop: PRIORITIZE → IMPLEMENT → REVIEW → MERGE in waves |
parallel-implement.sh |
N agents implementing issues in parallel via git worktrees |
parallel-review.sh |
Review all open MRs/PRs concurrently |
sequential-merge.sh |
Smart merge chain with rebase ordering |
| Rule | Purpose |
|---|---|
00-governance |
Governance baseline |
01-investigation-before-implementation |
Investigate before coding |
03-review-gate |
No completion without review |
04-tdd-mandatory |
TDD for behavior changes |
06-vcs-policy |
VCS conventions |
09-file-operations |
Safe file operations |
25-clean-code |
Clean code standards |
26-incremental-changes |
Small, focused changes |
27-commit-messages |
Semantic commit format |
30-posttask |
Mandatory final validation |
Java/Spring Boot: hexagonal architecture, clean code, layering, DTO design, MapStruct, Flyway, PostgreSQL, specifications, tenant isolation
TypeScript/NestJS: hexagonal architecture, AI orchestration, TDD, enums, database safety
Frontend: Feature-Sliced Design, Next.js, React, TypeScript
Security & Observability: data security, observability, redaction
Process: plans, docs, UX, changelog, MCP
Monorepo: submodule governance, merge/squash/CI-skip
| Agent | Role |
|---|---|
backend |
Backend implementation, APIs, persistence, integrations |
frontend |
UI, routing, state management, typed API consumption |
code-reviewer |
Findings-first technical review gate |
verifier |
Validation gate — posttask execution and evidence |
| Agent | Role |
|---|---|
arch |
Architecture decisions, boundary validation, ADRs |
ai-orchestrator |
AI/LLM orchestration, prompt contracts, model routing |
code-audit |
Repository-wide quality auditor |
docs |
Documentation governance |
docs-audit |
Documentation quality auditor, staleness detection |
project-manager |
Issue/milestone orchestration, plan-to-issue sync |
ux-reviewer |
UX quality gate for user-facing flows |
| Tier | Price | What you get |
|---|---|---|
| Free | $0 | 14 skills, 10 rules, 4 agents — essential governance |
| Pro | $79 | 67 skills, 51 rules, 11 agents, full pipeline, workflow docs, 22 design patterns |
| Team | $199 | Pro + 1 year updates + priority GitHub Issues support |
| Enterprise | $499/yr | Team + 2h setup consultancy + SLA |
| Add-on | Price | What you get |
|---|---|---|
| CI/CD Pipeline Pack | $29 | GitLab CI templates + GitHub Actions + AI changelog + version bump |
| Deploy Scripts Pack | $29 | Docker Swarm deploy, rollback, backup/restore, healthcheck |
| Full Bundle | $149 | Pro + CI/CD + Deploy (save $38) |
| Feature | AgentKit | Devin | Cursor | Codegen | Sweep |
|---|---|---|---|---|---|
| Governance framework | ✅ | ❌ | ❌ | ❌ | ❌ |
| Quality gates | 5 | 0 | 0 | 0 | 0 |
| TDD enforcement | ✅ | ❌ | ❌ | ❌ | ❌ |
| Multi-agent review chain | ✅ | ❌ | ❌ | ❌ | ❌ |
| Parallel via worktrees | ✅ | ❌ | ❌ | ❌ | ❌ |
| Autonomous pipeline loop | ✅ | ❌ | ❌ | ❌ | ❌ |
| Monorepo-aware | ✅ | ❌ | Partial | ❌ | ❌ |
| 22 GoF design patterns | ✅ | ❌ | ❌ | ❌ | ❌ |
| GitHub + GitLab | ✅ | GitHub | ❌ | GitHub | GitHub |
| Open source core | ✅ | ❌ | ❌ | ❌ | ✅ |
See CONTRIBUTING.md for guidelines on creating skills, rules, and agents.
MIT — see LICENSE for details.
The free tier is fully open source. Pro tier is distributed as a paid package.