AI Engineering System FoundationScaffolding, not lock-in
Discord: 3,800+ builders

You don't need our system.You need yours.

AgentKit gives you battle-tested agents, skills and workflows so you don't have to build your AI development system from zero. Keep what works. Change what doesn't. Make it yours.

Choose Your Starting PointSee How AgentKit Works
Foundational kits & tools:AgentKit EngineerAgentKit MarketingAgentKit BundleAgentKit App

Join hundreds of happy builders using AgentKit

3,800+ on Discord
System Composition
Built to Extend

From Battle-Tested Foundation to Your AI System

01AgentKit Foundation

Starting Point

Battle-tested agents, skills, hooks & adapters

02Your Conventions & Stack

Composable

Your project rules, tooling, and coding agents

03Your AI Engineering System

Full Ownership

Autonomous delivery workflows that fit how you build

Battle-Tested Scaffolding + Full OwnershipNo Lock-in
The real bottleneck

Tired of Building AI Workflows from Scratch for Claude Code?

You know coding agents are powerful, but reinventing context, conventions, and workflows across every project is slow and inconsistent.

Starting from Zero

Spending weeks discovering prompts, subagent roles, and tool integrations through trial and error.

Context & Quality Drift

Every session feels like starting over with different code styles, missed requirements, and fragmented standards.

Framework Lock-in

Rigid platforms that force you into their opinionated cage instead of adapting to your team’s workflow.

You don’t need another opinionated framework. You need a battle-tested starting point you can customize and own.

The architecture

AgentKit: A Battle-Tested Foundation for Your AI System

Connect specialized subagents, deterministic skills, and safety hooks into a delivery engine tailored to Claude Code.

Scaffolding, not vendor lock-in.

Build an AI system that matches how you write software

Keep what works. Change what doesn't. Replace it. Delete it. Extend it.

1Generic Coding Agent
2AgentKit Foundation
3Your Conventions & Tools
4Your AI Engineering System
108+

Skills

Pre-built skills for engineering, research, code review, and automation

95+

Commands

Slash commands covering the full delivery lifecycle

45

Subagents

Specialized role agents ready to compose into custom workflows

01 / Core Architecture

Four Pillars of Production Agentics

01Focused Roles

Specialized Subagents

Role-based delegation

Dedicated subagents for planning, coding, code review, security audits, and deployment.

02Delivery Pipeline

Chained Skill Graph

Guarded delivery loop

Skills connect into structured pipelines (Brainstorm → Plan → Cook → Review → Ship) with deterministic gates.

03Active Safety

Context & Safety Hooks

Just-in-time guardrails

Pre-tool hooks prompt for approval before sensitive operations and inject project rules at key execution points.

04Multi-Harness

Harness-Native Adapters

Deep runtime integration

Deeply adapted to Claude Code, Codex, Cursor with native tool definitions and model routing.

02 / Chained Workflows

Interactive Skill Graph

Skills trigger skills automatically — passing evidence, contracts, and review state across stages.

  1. STAGE 01/ak:brainstorm

    Explore & Shape Intent

    Exhaust trade-offs, identify edge cases, and define non-goals before touching code.

    Brainstormer Agent
  2. STAGE 02/ak:advise

    Advisory Consultation

    Autonomous Fable-tier counsel reframes requirements, surfaces anti-patterns, and scores options.

    Kongming SupervisorFable 5 Supervised
  3. STAGE 03/ak:plan

    5-Phase Architecture

    Generates verifiable phase contracts, dependency DAGs, and interactive HTML artifacts.

    Planner Specialist
  4. STAGE 04/ak:cook

    Structured Implementation

    TDD tests-first execution, worktree isolation, and zero-breakage modularization.

    Fullstack Dev Agent
  5. STAGE 05/ak:code-review

    Evidence-Based Review

    Multi-perspective quality audit, regression tests, and security scans.

    Code Reviewer
  6. STAGE 06/ak:ship

    Guarded Merge & CI Watch

    Conventional commit splitting, PR verification, and CI convergence watch.

    Git Manager
03 / Safety & Invariants

Lifecycle Hooks & Pre-Tool Safety Gates

AgentKit injects rules and context at key execution points to guide agent execution safely.

01. Pre-Session

Context Injection & Rules

Auto-loads AGENTS.md, project invariants, and skills index into agent harness.

02. Pre-Tool ExecutionActive Gate

Pre-Tool Approval & Checks

Prompts for approval on sensitive tools and checks parameters before execution.

03. Post-Tool Execution

Output Capture & Tracing

Captures tool outputs, stores structured artifacts, and records execution traces.

04. Pre-Commit

Code Review & Lint Check

Validates commit messages and runs targeted test assertions before commit.

05. Merge Train

Guarded Merge Supervised

Serializes production merge queue and checks CI to green.

06. Post-Merge Watch

Release & Deploy Verification

Monitors production health checks and telemetry baselines.

PRE-TOOL SAFETY CHECKS & PROMPTS● SAFETY CHECKS
✓DEFAULT PERMITTED OPERATIONS
✓Scoped directory reads
✓Unit test executions
✓Non-destructive dry runs
✓AST-aware surgical edits
✕APPROVAL REQUIRED ON RISK
✕Force push to protected branches
✕Raw secrets in terminal logs
✕Destructive DROP / TRUNCATE DDL
✕Unverified external network sends
Enforcement mechanisms and hook coverage vary by coding agent harness.
04 / Multi-Harness Portability

Harness Compatibility & Support Tiers

Truth in marketing: We clearly distinguish production-supported coding agents from experimental preview builds and planned integrations.

TierSupported HarnessesIntegration Status
Production
Native Production Support
Claude CodeCodexCursor

Official runtime adapter with native tool dispatch and model routing.

Preview
Experimental Preview
AntigravityOpenCodeGrokGitHub Copilot

Early-access runtime adapter undergoing active testing and feedback.

Planned
Roadmap Target
Oh My PiPiDroidWindsurfRoo CodeKilo CodeAmpClineOpenHandsOpenClawGoClawDewee

Scheduled for future adapter development.

✓ Instant CLI setup•✓ Works with your stack•✓ Full ownership & customization

Start building with a battle-tested foundation today. Adapt everything as your team evolves.

Foundational Kits

Choose Your Starting Foundation

Start with pre-built kits and tools, then customize every agent, skill, and workflow to match your team.

Engineering

AgentKit Engineer

For software engineers & full-stack builders

$99one-time kit access

60+ skills, 30+ workflows, and AI subagents for fullstack software engineering

  • TDD planning and execution
  • Database & API development
  • DevOps & CI/CD automation
  • Evidence-based code review
Explore AgentKit Engineer
Full Foundation

AgentKit Bundle

Full suite: Engineer + Marketing with multi-agent orchestration

$149(Save $49 vs $198 separately)

The complete foundation: AgentKit Engineer + AgentKit Marketing with priority updates

  • All 108+ engineering & marketing skills
  • All 45 specialized subagents
  • 3-device Desktop App license included
  • Priority workflow updates & Discord access
Get Complete Foundation
Growth & Sales

AgentKit Marketing

For growth teams, founders & marketers

$99one-time kit access

12+ MCP integrations and growth workflows for research, sales, and content

  • Automated research & brief creation
  • Programmatic SEO & content pipeline
  • Competitor & market intelligence
  • Multi-channel copywriting agents
Explore AgentKit Marketing

AgentKit App

Desktop CockpitmacOS & Windows

Desktop license manager and visual cockpit for inspecting and tuning your system

$49/yr (1 device)•$99/yr (3 devices)•Companion cockpit · Kit licenses sold separately
Get Desktop License
Visual Cockpit

One Interface to Inspect & Manage Your System

Use the AgentKit Desktop App as a companion cockpit to inspect your Skill Graph, manage licensed kits, and configure status lines for Claude Code.

  • Inspect your Skill Graph and active subagents visually
  • Manage licensed kits and CLI integrations in one click
  • Configure status lines and safety hooks across your projects

Desktop App Licenses

Instant license key delivery. Requires Claude Code, Codex, or OpenAI account.

1 Device License

Full Desktop App access for 1 workstation.

$49/year

Renews annually. 14-day money-back guarantee.

3 Devices License

Desktop App access for up to 3 workstations with priority updates.

Most Popular
$99/year

Best for developers working across desktop and laptop.

Explore Desktop Cockpit
Design your agent status line visually
AgentKit Desktop App status line builder with live preview
Activate and manage licensed kits
AgentKit Desktop App kits screen for activating licensed kits
Browse 200+ installed skills
AgentKit Desktop App skills browser listing installed skills
Proof

What Customers Are Saying

Real feedback from developers using AgentKit in production

4.9/5 from 349 ratings · 12 customer reviews

Join hundreds of happy builders using AgentKit

3,800+ on Discord
Discord testimonial from Hoan Nguyen

“Tried @Goon's .claude file from the human-mcp repo and it's really impressive. Need to test it more to truly appreciate it. Hope this helps everyone.”

Discord testimonial from Philonik

“You're amazing.”

Facebook comment from Lương Thanh Thế

“Duy Nguyen has AgentKit. I find it much more stable, the whole project doesn't break. Currently my 20x package is limited because I need to review everything carefully, I don't know why Claude AI has developed so many unnecessary things, and many duplicates, but I don't know how to fix it either. So having your kit is like hitting gold.”

Discord testimonial from Zachary Nguyen

“Anyway, been playing with AgentKit all day, so far so good! The /plan and /cook feature, very impressive 😄”

Discord testimonial from cesc

“No kit yet, just 8 hours with kit, now 16 hours =)))”

Message with CodeRabbit screenshot

“Confirming it's super smooth, it fixed all the CodeRabbit issues it found. Worth the $30/month, big time!”

X.com post from LAM (@laingoclam)

“I completed the feature for iamhere.to with just a single prompt. #AgentKit by @goon_nguyen is truly effective.”

Discord testimonial from Zachary Nguyen

“This speed is really impressive, first time seeing it so calm again.”

Discord testimonial from Hải Đăng

“Using AI tools, bought this kit, now understanding how convenient it is for coders. Opening the project and running 2-3 Claude instances doing different features but still feeling confident.”

Discord testimonial from Mr The

“Cooking all day and night. Admit there's no plan, cook, test always, Claude Code is really easy to get confused. Plus the docs are aligned with everything. Before telling Claude Code it would take a while to align the project, but only within 1 session it's all in. Now having AgentKit is really convenient.”

X.com post from Florin Makes (@florinmakes)

“I really like its structured process with specialized agents. Really considering upgrading to max because the process with this is amazing.”

Discord testimonial from phuongdeptrai

“Used AgentKit for 2 days and got cursed by colleagues. Specifically, too free so got cursed 🤣”

Ecosystem

Other Products

Explore our ecosystem of AI-powered tools

GoClaw logo

GoClaw(opens in new tab)

AI Agent Platform for Enterprises

Deploy, manage, and scale AI agents across your organization with enterprise-grade security and observability.

Dewee logo

Dewee(opens in new tab)

AI workspace for focused builders

Plan, build, and operate projects from a focused AI-native workspace for serious shipping.

Cloud Harness MCP logo

Cloud Harness MCP(opens in new tab)

Remote Coding Harness via MCP

Authenticated remote Docker coding workspaces, Git & task automation for trusted operators.

AgentBrain logo

AgentBrain(opens in new tab)

AI-Powered Data Management

Multi-source connectors, pgvector semantic search, and automated visual ETL pipelines.

AgentWiki logo

AgentWiki(opens in new tab)

Knowledge Platform for Humans & AI

Full-stack knowledge management on Cloudflare with MCP tools, hybrid search, and knowledge graph.

TOSE logo

TOSE(opens in new tab)

Infrastructure Automation Platform

Git-to-Kubernetes automated container builds, deployments, real-time logs, and multi-tenant teams.

Don’t build your AI system from zero. Make it yours. Get started today.

Join hundreds of happy builders using AgentKit

3,800+ on Discord
Explore AgentKit Engineer — $99Explore AgentKit Marketing — $99
Join 3,800+ builders on Discord
14-day refund guarantee
Instant license-key access
Full source code ownership
Loading blog articles...