v3.0.0Phi Công

Product Timeline

The journey of Flezi AgentBox — from first commit to production platform. Every version tells a story.

187

Epics

747

Features

4,394

Tests

Days

7 tháng 4, 2026

v3.0.0Current

Phi Công

The Autopilot

Agent Autopilot — background daemon orchestrates agents automatically. Event routing, priority queues, budget control, multi-agent pipelines, PR review bot, and approval workflows.

Autopilot Daemon — agentbox autopilot on — background daemon orchestrates watch, hooks, cron. One command to activate your AI team.
Event Router + Execution Queue — Match file_save/pre_commit/pr_create/cron events to agents. Priority heap queue with max 2 concurrent, rate limiting.
Multi-Agent Pipelines — Chain agents sequentially (A → B → C) or swarm in parallel. Event bus with file-based pub/sub.
PR Review Bot — GitHub/GitLab webhook → multi-agent parallel review → inline comments with fix suggestions.
Budget & Model Switching — Daily/weekly USD limits. Auto-degrade Cloud → Mid-Cloud → Ollama on budget pressure. Priority agent protection.
Approval Workflow + Trust Gate — T0-T1 agents queue for approval. T2 auto-fix. T3 full autopilot. Weekly digest reports.

Epics

Features

4,394

Tests

Epics 177–187

v3.0.0Current

Phi Công

The Autopilot

Agent Autopilot — background daemon orchestrates agents automatically. Event routing, priority queues, budget control, multi-agent pipelines, PR review bot, and approval workflows.

Autopilot Daemon — agentbox autopilot on — background daemon orchestrates watch, hooks, cron. One command to activate your AI team.
Event Router + Execution Queue — Match file_save/pre_commit/pr_create/cron events to agents. Priority heap queue with max 2 concurrent, rate limiting.
Multi-Agent Pipelines — Chain agents sequentially (A → B → C) or swarm in parallel. Event bus with file-based pub/sub.
PR Review Bot — GitHub/GitLab webhook → multi-agent parallel review → inline comments with fix suggestions.
Budget & Model Switching — Daily/weekly USD limits. Auto-degrade Cloud → Mid-Cloud → Ollama on budget pressure. Priority agent protection.
Approval Workflow + Trust Gate — T0-T1 agents queue for approval. T2 auto-fix. T3 full autopilot. Weekly digest reports.

Epics

Features

4,394

Tests

Epics 177–187

7 tháng 4, 2026

3 tháng 4, 2026

v2.4.0

Trải Nghiệm

The Experience

Agent Detail UX — interactive playground, before/after showcase, effectiveness metrics, onboarding wizard, dark mode.

Interactive Playground — Try agents in browser — pre-loaded samples, paste own code, scan results inline.
Before/After Showcase — See vulnerable code → agent fix with syntax-highlighted diffs per agent.
Onboarding Wizard — 3-step: pick tech stack → recommended agents → first scan. Value in 60 seconds.

Epics

Features

4,046

Tests

Epics 173–176

3 tháng 4, 2026

v2.4.0

Trải Nghiệm

The Experience

Agent Detail UX — interactive playground, before/after showcase, effectiveness metrics, onboarding wizard, dark mode.

Interactive Playground — Try agents in browser — pre-loaded samples, paste own code, scan results inline.
Before/After Showcase — See vulnerable code → agent fix with syntax-highlighted diffs per agent.
Onboarding Wizard — 3-step: pick tech stack → recommended agents → first scan. Value in 60 seconds.

Epics

Features

4,046

Tests

Epics 173–176

3 tháng 4, 2026

2 tháng 4, 2026

v1.9.0

Trí Nhớ

The Memory

Agent Memory & Personalization — 3-layer persistent memory, maturity L1-L5, trust escalation T0-T3.

3-Layer Memory — Global (personal) + Project (team-shared) + Session. Preferences, feedback, learnings.
Maturity L1-L5 — New → Learning → Calibrated → Expert → Veteran. Agents grow with use.
Trust Escalation T0-T3 — Earned via track record: sessions + acceptance rate. More trust → more autonomy.

Epics

Features

3,994

Tests

Epics 163–164

v1.9.0

Trí Nhớ

The Memory

Agent Memory & Personalization — 3-layer persistent memory, maturity L1-L5, trust escalation T0-T3.

3-Layer Memory — Global (personal) + Project (team-shared) + Session. Preferences, feedback, learnings.
Maturity L1-L5 — New → Learning → Calibrated → Expert → Veteran. Agents grow with use.
Trust Escalation T0-T3 — Earned via track record: sessions + acceptance rate. More trust → more autonomy.

Epics

Features

3,994

Tests

Epics 163–164

2 tháng 4, 2026

v2.0.0

Tự Chủ

The Autonomy

Autonomous Workflows — watch daemon, YAML workflow engine, pre-commit 2-tier scanning, multi-agent pipelines.

Watch Daemon — agentbox watch — react to file changes, trigger workflows automatically.
Pre-commit 2-Tier Scan — Tier 1 regex (<1s) + Tier 2 heuristic. Block secrets, SQLi, CMDi before commit.
Workflow YAML Engine — Declarative steps, conditions, budget control. Like GitHub Actions for AI agents.

Epics

Features

4,030

Tests

Epics 165–168

2 tháng 4, 2026

v2.0.0

Tự Chủ

The Autonomy

Autonomous Workflows — watch daemon, YAML workflow engine, pre-commit 2-tier scanning, multi-agent pipelines.

Watch Daemon — agentbox watch — react to file changes, trigger workflows automatically.
Pre-commit 2-Tier Scan — Tier 1 regex (<1s) + Tier 2 heuristic. Block secrets, SQLi, CMDi before commit.
Workflow YAML Engine — Declarative steps, conditions, budget control. Like GitHub Actions for AI agents.

Epics

Features

4,030

Tests

Epics 165–168

2 tháng 4, 2026

v2.1.0

Tiến Hoá

The Evolution

Agent Factory + Self-Update — auto-generate agents from descriptions, self-evolve from feedback, A/B testing.

Agent Factory — Describe in natural language → 6-stage pipeline generates agent with persona + knowledge + policies.
Self-Update — Agents analyze feedback → modify knowledge → A/B test → auto-version bump.
20 Demo Agents — Full catalog: security, quality, testing, performance, docs, style, CI, i18n, changelog.

Epics

Features

4,046

Tests

Epics 169–172

v2.1.0

Tiến Hoá

The Evolution

Agent Factory + Self-Update — auto-generate agents from descriptions, self-evolve from feedback, A/B testing.

Agent Factory — Describe in natural language → 6-stage pipeline generates agent with persona + knowledge + policies.
Self-Update — Agents analyze feedback → modify knowledge → A/B test → auto-version bump.
20 Demo Agents — Full catalog: security, quality, testing, performance, docs, style, CI, i18n, changelog.

Epics

Features

4,046

Tests

Epics 169–172

2 tháng 4, 2026

1 tháng 4, 2026

v1.7.0

Nhận Thức

The Awareness

Context-Aware Execution — agents auto-detect project type, language, framework, and dependencies. Zero config.

5 Language Detectors — Go, Node/TypeScript, Python, Rust, Java — auto-detected from signal files
Monorepo Detection — Turborepo, pnpm, Nx, Lerna — detect root + sub-project context
Token-Budgeted Injection — Max 2000 tokens context injected into agent prompt. Cache <10ms.

Epics

Features

3,869

Tests

Epics 158–159

1 tháng 4, 2026

v1.7.0

Nhận Thức

The Awareness

Context-Aware Execution — agents auto-detect project type, language, framework, and dependencies. Zero config.

5 Language Detectors — Go, Node/TypeScript, Python, Rust, Java — auto-detected from signal files
Monorepo Detection — Turborepo, pnpm, Nx, Lerna — detect root + sub-project context
Token-Budgeted Injection — Max 2000 tokens context injected into agent prompt. Cache <10ms.

Epics

Features

3,869

Tests

Epics 158–159

1 tháng 4, 2026

v1.8.0

Đơn Giản

The Simplicity

CLI UX Redesign — Docker/npx-style agentbox <agent> shorthand, BMAD-style override layer, project-local agents.

agentbox <agent> Shorthand — 1 command to run any agent. Auto-pull, one-shot, pipe, --serve modes.
Override Layer (BMAD-style) — .base/ + overrides/ + knowledge/ — customize agents per-project, team share via git.
.agentbox/config.yaml — npm-like manifest: agents, default, shortcuts, shared_knowledge. agentbox install after clone.

Epics

Features

3,950

Tests

Epics 160–162

v1.8.0

Đơn Giản

The Simplicity

CLI UX Redesign — Docker/npx-style agentbox <agent> shorthand, BMAD-style override layer, project-local agents.

agentbox <agent> Shorthand — 1 command to run any agent. Auto-pull, one-shot, pipe, --serve modes.
Override Layer (BMAD-style) — .base/ + overrides/ + knowledge/ — customize agents per-project, team share via git.
.agentbox/config.yaml — npm-like manifest: agents, default, shortcuts, shared_knowledge. agentbox install after clone.

Epics

Features

3,950

Tests

Epics 160–162

1 tháng 4, 2026

31 tháng 3, 2026

v1.6.0

Kết Nối

The Bridge

Universal MCP Local Server — agents accessible from any IDE, Slack bot, CI/CD, or script via stdio + HTTP dual transport.

MCP Server (stdio + HTTP) — agentbox mcp serve hosts agents as tools for Claude Code, Cursor, VS Code, and any MCP client
Agent-as-Tool Pattern — ask_{name}(question) → persona + knowledge + LLM → response. Hot reload on pull.
Agent Mesh — Agents call agents internally via MCP routing. Multi-agent collaboration.

Epics

Features

3,825

Tests

Epics 155–157

31 tháng 3, 2026

v1.6.0

Kết Nối

The Bridge

Universal MCP Local Server — agents accessible from any IDE, Slack bot, CI/CD, or script via stdio + HTTP dual transport.

MCP Server (stdio + HTTP) — agentbox mcp serve hosts agents as tools for Claude Code, Cursor, VS Code, and any MCP client
Agent-as-Tool Pattern — ask_{name}(question) → persona + knowledge + LLM → response. Hot reload on pull.
Agent Mesh — Agents call agents internally via MCP routing. Multi-agent collaboration.

Epics

Features

3,825

Tests

Epics 155–157

31 tháng 3, 2026

30 tháng 3, 2026

v1.5.0

Sẵn Sàng

The Readiness

CLI Production Hardening — E2E demo, agent package format, MCP tool-calling engine, rate limit retry, Windows support.

CLI E2E Demo — 19-step end-to-end demo: init → pull → run → scan → report. Fully automated.
MCP Tool-Calling Engine — Local tool execution, per-agent MCP config, connected mode with server-side tools/call.
Agent Package Format — .agentbox format: agent.yaml v2 + persona.md + knowledge/ + tools/ + policies/
Production Hardening — Rate limit retry, Windows support, graceful shutdown, NemoClaw mandatory enforcement.

Epics

Features

3,810

Tests

Epics 136–154

v1.5.0

Sẵn Sàng

The Readiness

CLI Production Hardening — E2E demo, agent package format, MCP tool-calling engine, rate limit retry, Windows support.

CLI E2E Demo — 19-step end-to-end demo: init → pull → run → scan → report. Fully automated.
MCP Tool-Calling Engine — Local tool execution, per-agent MCP config, connected mode with server-side tools/call.
Agent Package Format — .agentbox format: agent.yaml v2 + persona.md + knowledge/ + tools/ + policies/
Production Hardening — Rate limit retry, Windows support, graceful shutdown, NemoClaw mandatory enforcement.

Epics

Features

3,810

Tests

Epics 136–154

30 tháng 3, 2026

25 tháng 3, 2026

v1.1.0

Vuốt Sắc

The Claw

NemoClaw mandatory integration — kernel-level agent sandboxing with Landlock + seccomp + netns. L0-L3 permission model, binary integrity verification.

NemoClaw Sandbox Runtime — Kernel-isolated execution with Landlock + seccomp + network namespaces. No fallback — mandatory for all agents.
L0–L3 Permission Model — 4-tier trust: Untrusted → Community → Verified → Enterprise. Kernel-enforced restrictions per tier.
Binary Integrity — SHA-256 hash verification on every NemoClaw startup. Tamper-proof agent isolation.
IPC Protocol — JSONL over stdio pipe between CLI and NemoClaw child process. Cross-platform: native Linux, Docker fallback Win/macOS.

Epics

Features

3,785

Tests

Epics 130–135

25 tháng 3, 2026

v1.1.0

Vuốt Sắc

The Claw

NemoClaw mandatory integration — kernel-level agent sandboxing with Landlock + seccomp + netns. L0-L3 permission model, binary integrity verification.

NemoClaw Sandbox Runtime — Kernel-isolated execution with Landlock + seccomp + network namespaces. No fallback — mandatory for all agents.
L0–L3 Permission Model — 4-tier trust: Untrusted → Community → Verified → Enterprise. Kernel-enforced restrictions per tier.
Binary Integrity — SHA-256 hash verification on every NemoClaw startup. Tamper-proof agent isolation.
IPC Protocol — JSONL over stdio pipe between CLI and NemoClaw child process. Cross-platform: native Linux, Docker fallback Win/macOS.

Epics

Features

3,785

Tests

Epics 130–135

25 tháng 3, 2026

24 tháng 3, 2026

v1.0.0

Khởi Đầu

The Beginning

v1.0 release — observability, production deployment, agent runtime pipeline with Claude Code CLI, BYOK key vault, runtime configuration, and user journey demo flows.

Agent Runtime Pipeline — RuntimeFactory with Docker and Claude Code CLI/SDK runtimes, MCP server config
BYOK Key Vault — Bring Your Own Key — AES-256-GCM encrypted storage for Anthropic, OpenAI, and Google API keys
Runtime Configuration — Per-agent model, budget, timeout, tool permissions, and MCP server settings in Builder Lab
Production Deployment — Docker Compose production stack with Caddy reverse proxy and health checks
Performance & Observability — k6 load tests, Sentry, structured logging, Redis caching, Lighthouse CI

Epics

141

Features

3,248

Tests

Epics 83–129

v1.0.0

Khởi Đầu

The Beginning

v1.0 release — observability, production deployment, agent runtime pipeline with Claude Code CLI, BYOK key vault, runtime configuration, and user journey demo flows.

Agent Runtime Pipeline — RuntimeFactory with Docker and Claude Code CLI/SDK runtimes, MCP server config
BYOK Key Vault — Bring Your Own Key — AES-256-GCM encrypted storage for Anthropic, OpenAI, and Google API keys
Runtime Configuration — Per-agent model, budget, timeout, tool permissions, and MCP server settings in Builder Lab
Production Deployment — Docker Compose production stack with Caddy reverse proxy and health checks
Performance & Observability — k6 load tests, Sentry, structured logging, Redis caching, Lighthouse CI

Epics

141

Features

3,248

Tests

Epics 83–129

24 tháng 3, 2026

17 tháng 3, 2026

v0.9.0

Bình Minh

The Dawn

Enterprise-ready — team workspaces, real-time collaboration, horizontal scaling, Storybook, and code review hardening.

Team Workspaces — Organizations with role-based access, invites, and workspace switching
Real-time Collaboration — Live presence indicators and concurrent editing support
Security Hardening — 41 fixes from adversarial code review (8C/13H/20M/16L findings)
Storybook — Component library documentation with Chromatic visual testing

Epics

Features

2,052

Tests

Epics 61–82

17 tháng 3, 2026

v0.9.0

Bình Minh

The Dawn

Enterprise-ready — team workspaces, real-time collaboration, horizontal scaling, Storybook, and code review hardening.

Team Workspaces — Organizations with role-based access, invites, and workspace switching
Real-time Collaboration — Live presence indicators and concurrent editing support
Security Hardening — 41 fixes from adversarial code review (8C/13H/20M/16L findings)
Storybook — Component library documentation with Chromatic visual testing

Epics

Features

2,052

Tests

Epics 61–82

17 tháng 3, 2026

14 tháng 3, 2026

v0.7.0

Cánh Buồm

The Sails

User ecosystem — tabbed agent detail pages, creator profiles, documentation hub, feedback system, Go CLI, and distribution.

Agent Detail Pages — Tabbed interface: Overview, Reviews, API, Changelog, MCP
Creator Storefronts — Public creator profiles with portfolio and earnings dashboard
Documentation Hub — MDX-powered docs with search, code highlighting, and API reference
Go CLI — Full-featured CLI with Cobra/Viper for agent management

Epics

Features

1,660

Tests

Epics 43–60

v0.7.0

Cánh Buồm

The Sails

User ecosystem — tabbed agent detail pages, creator profiles, documentation hub, feedback system, Go CLI, and distribution.

Agent Detail Pages — Tabbed interface: Overview, Reviews, API, Changelog, MCP
Creator Storefronts — Public creator profiles with portfolio and earnings dashboard
Documentation Hub — MDX-powered docs with search, code highlighting, and API reference
Go CLI — Full-featured CLI with Cobra/Viper for agent management

Epics

Features

1,660

Tests

Epics 43–60

14 tháng 3, 2026

11 tháng 3, 2026

v0.5.0

Nền Tảng

The Foundation

Advanced platform — multi-LLM support, agent templates, versioning, pipeline DAG editor, marketplace economics, and OWASP security hardening.

Multi-LLM Support — Claude, GPT-4, Gemini with model selection and fallback
Agent Templates — Pre-built templates for common use cases with one-click cloning
Pipeline DAG Editor — Visual workflow builder with Action, Service, and Gate nodes
Marketplace Economics — Creator revenue sharing, pricing tiers, and analytics
OWASP Hardening — 54/54 security checks passing with CSRF, XSS, SQLi protection

Epics

Features

1,200

Tests

Epics 20–42

11 tháng 3, 2026

v0.5.0

Nền Tảng

The Foundation

Advanced platform — multi-LLM support, agent templates, versioning, pipeline DAG editor, marketplace economics, and OWASP security hardening.

Multi-LLM Support — Claude, GPT-4, Gemini with model selection and fallback
Agent Templates — Pre-built templates for common use cases with one-click cloning
Pipeline DAG Editor — Visual workflow builder with Action, Service, and Gate nodes
Marketplace Economics — Creator revenue sharing, pricing tiers, and analytics
OWASP Hardening — 54/54 security checks passing with CSRF, XSS, SQLi protection

Epics

Features

1,200

Tests

Epics 20–42

11 tháng 3, 2026

8 tháng 3, 2026

v0.3.0

Rễ Cây

The Roots

Quality infrastructure — E2E test suite, CI/CD pipelines, Shadcn/UI migration, AppShell, dark mode, and accessibility compliance.

E2E Test Suite — Playwright tests with page objects, fixtures, and CI sharding
Shadcn/UI Migration — Complete component library migration with design tokens
Dark Mode — System-aware theme switching with CSS custom properties
Accessibility — WCAG 2.1 AA compliance with axe-core automated checks

Epics

Features

480

Tests

Epics 8–19

v0.3.0

Rễ Cây

The Roots

Quality infrastructure — E2E test suite, CI/CD pipelines, Shadcn/UI migration, AppShell, dark mode, and accessibility compliance.

E2E Test Suite — Playwright tests with page objects, fixtures, and CI sharding
Shadcn/UI Migration — Complete component library migration with design tokens
Dark Mode — System-aware theme switching with CSS custom properties
Accessibility — WCAG 2.1 AA compliance with axe-core automated checks

Epics

Features

480

Tests

Epics 8–19

8 tháng 3, 2026

6 tháng 3, 2026

v0.1.0

Hạt Giống

The Seed

Foundation release — authentication, marketplace browsing, agent builder lab, Docker execution engine, and admin dashboard.

Auth & User System — Supabase Auth with email/password, RLS policies, session management
Agent Marketplace — Browse, search, filter agents with category and tag system
Builder Lab — Three-phase guided workflow: Discovery → Logic → Output
Execution Sandbox — Docker-isolated agent execution with SSE streaming
MCP Integration — Model Context Protocol support for IDE connections

Epics

Features

120

Tests

Epics 1–7

6 tháng 3, 2026

v0.1.0

Hạt Giống

The Seed

Foundation release — authentication, marketplace browsing, agent builder lab, Docker execution engine, and admin dashboard.

Auth & User System — Supabase Auth with email/password, RLS policies, session management
Agent Marketplace — Browse, search, filter agents with category and tag system
Builder Lab — Three-phase guided workflow: Discovery → Logic → Output
Execution Sandbox — Docker-isolated agent execution with SSE streaming
MCP Integration — Model Context Protocol support for IDE connections

Epics

Features

120

Tests

Epics 1–7

6 tháng 3, 2026

Next-Gen Agent Platform

v4.0.0

Vô Hạn

— The Infinite

Multi-modal agents — code + vision + voice in one conversation.

v4.0 takes AgentBox beyond code to multi-modal intelligence. Agents that understand screenshots, design mockups, and voice commands. Federated agent networks across organizations. Real-time human-agent collaboration. The future of AI-powered software development.

Target: Q3 2026

Agents Built-in

187

Epics Shipped

4,394

Total Tests

Autopilot Presets

CLI Packages

Event Types

Universal Runtime — NemoClawLite

Kernel-level isolation (Landlock + seccomp + netns) for every tier. Same security, different features.

Free

NemoClawLite + Local LLM

Full CLI, kernel isolation, Ollama local LLM, unlimited agents, 20 built-in agents, autopilot

Popular

Pro

$19/month

NemoClawLite + Cloud LLM

Cloud hosting, team features, Claude / GPT integration, memory sync, PR review bot, priority support

Enterprise

NemoClawLite + Custom LLM

Full policy control, audit trail, SSO, compliance, private registry, mission control dashboard, SLA

What's Coming in v2.0

6 major features across ecosystem, intelligence, security, and enterprise

Multi-Modal Agents

IntegrationPlanned

Agents that understand code, screenshots, design mockups, and voice commands. Vision + code + text in one conversation.

Federated Agent Networks

PlatformResearch

Cross-organization agent collaboration. Shared knowledge without sharing code. Privacy-preserving federated learning.

Real-Time Human-Agent Collab

IntelligencePlanned

Live pair programming with agents. Shared cursor, inline suggestions, voice commands. Google Docs for code.

Agent Marketplace v2

EcosystemPlanned

Workflow marketplace: sell curated agent packs + triggers + config as bundles. Recurring revenue model.

Mobile Companion PWA

EnterprisePlanned

Push notifications, approval swipe, daily summary cards. Monitor autopilot on the go.

Self-Hosted Air-Gapped

SecurityPlanned

On-premise deployment for regulated industries. Local model inference, no external calls.

Development Progress

Track every epic across v1.0, v1.1, and the upcoming v2.0 Agent Ecosystem — with real-time progress bars and a visual Gantt chart.

View Gantt Chart & Progress

Beyond v2.0

What's on the horizon

Planned

Mobile Companion App

iOS and Android app for monitoring agent executions and quick interactions on the go

Planned

Self-Hosted On-Premise

Air-gapped deployment option for regulated industries with local model inference

Planned

Marketplace API

Public REST + GraphQL API for third-party integrations and custom agent storefronts

Planned

Multi-Tenant White Label

Resellable platform with custom branding, domains, and tenant-isolated infrastructure

v3.0.0Phi Công

Product Timeline

The journey of Flezi AgentBox — from first commit to production platform. Every version tells a story.

187

Epics

747

Features

4,394

Tests

Days

7 tháng 4, 2026

v3.0.0Current

Phi Công

The Autopilot

Agent Autopilot — background daemon orchestrates agents automatically. Event routing, priority queues, budget control, multi-agent pipelines, PR review bot, and approval workflows.

Autopilot Daemon — agentbox autopilot on — background daemon orchestrates watch, hooks, cron. One command to activate your AI team.
Event Router + Execution Queue — Match file_save/pre_commit/pr_create/cron events to agents. Priority heap queue with max 2 concurrent, rate limiting.
Multi-Agent Pipelines — Chain agents sequentially (A → B → C) or swarm in parallel. Event bus with file-based pub/sub.
PR Review Bot — GitHub/GitLab webhook → multi-agent parallel review → inline comments with fix suggestions.
Budget & Model Switching — Daily/weekly USD limits. Auto-degrade Cloud → Mid-Cloud → Ollama on budget pressure. Priority agent protection.
Approval Workflow + Trust Gate — T0-T1 agents queue for approval. T2 auto-fix. T3 full autopilot. Weekly digest reports.

Epics

Features

4,394

Tests

Epics 177–187

v3.0.0Current

Phi Công

The Autopilot

Agent Autopilot — background daemon orchestrates agents automatically. Event routing, priority queues, budget control, multi-agent pipelines, PR review bot, and approval workflows.

Autopilot Daemon — agentbox autopilot on — background daemon orchestrates watch, hooks, cron. One command to activate your AI team.
Event Router + Execution Queue — Match file_save/pre_commit/pr_create/cron events to agents. Priority heap queue with max 2 concurrent, rate limiting.
Multi-Agent Pipelines — Chain agents sequentially (A → B → C) or swarm in parallel. Event bus with file-based pub/sub.
PR Review Bot — GitHub/GitLab webhook → multi-agent parallel review → inline comments with fix suggestions.
Budget & Model Switching — Daily/weekly USD limits. Auto-degrade Cloud → Mid-Cloud → Ollama on budget pressure. Priority agent protection.
Approval Workflow + Trust Gate — T0-T1 agents queue for approval. T2 auto-fix. T3 full autopilot. Weekly digest reports.

Epics

Features

4,394

Tests

Epics 177–187

7 tháng 4, 2026

3 tháng 4, 2026

v2.4.0

Trải Nghiệm

The Experience

Agent Detail UX — interactive playground, before/after showcase, effectiveness metrics, onboarding wizard, dark mode.

Interactive Playground — Try agents in browser — pre-loaded samples, paste own code, scan results inline.
Before/After Showcase — See vulnerable code → agent fix with syntax-highlighted diffs per agent.
Onboarding Wizard — 3-step: pick tech stack → recommended agents → first scan. Value in 60 seconds.

Epics

Features

4,046

Tests

Epics 173–176

3 tháng 4, 2026

v2.4.0

Trải Nghiệm

The Experience

Agent Detail UX — interactive playground, before/after showcase, effectiveness metrics, onboarding wizard, dark mode.

Interactive Playground — Try agents in browser — pre-loaded samples, paste own code, scan results inline.
Before/After Showcase — See vulnerable code → agent fix with syntax-highlighted diffs per agent.
Onboarding Wizard — 3-step: pick tech stack → recommended agents → first scan. Value in 60 seconds.

Epics

Features

4,046

Tests

Epics 173–176

3 tháng 4, 2026

2 tháng 4, 2026

v1.9.0

Trí Nhớ

The Memory

Agent Memory & Personalization — 3-layer persistent memory, maturity L1-L5, trust escalation T0-T3.

3-Layer Memory — Global (personal) + Project (team-shared) + Session. Preferences, feedback, learnings.
Maturity L1-L5 — New → Learning → Calibrated → Expert → Veteran. Agents grow with use.
Trust Escalation T0-T3 — Earned via track record: sessions + acceptance rate. More trust → more autonomy.

Epics

Features

3,994

Tests

Epics 163–164

v1.9.0

Trí Nhớ

The Memory

Agent Memory & Personalization — 3-layer persistent memory, maturity L1-L5, trust escalation T0-T3.

3-Layer Memory — Global (personal) + Project (team-shared) + Session. Preferences, feedback, learnings.
Maturity L1-L5 — New → Learning → Calibrated → Expert → Veteran. Agents grow with use.
Trust Escalation T0-T3 — Earned via track record: sessions + acceptance rate. More trust → more autonomy.

Epics

Features

3,994

Tests

Epics 163–164

2 tháng 4, 2026

v2.0.0

Tự Chủ

The Autonomy

Autonomous Workflows — watch daemon, YAML workflow engine, pre-commit 2-tier scanning, multi-agent pipelines.

Watch Daemon — agentbox watch — react to file changes, trigger workflows automatically.
Pre-commit 2-Tier Scan — Tier 1 regex (<1s) + Tier 2 heuristic. Block secrets, SQLi, CMDi before commit.
Workflow YAML Engine — Declarative steps, conditions, budget control. Like GitHub Actions for AI agents.

Epics

Features

4,030

Tests

Epics 165–168

2 tháng 4, 2026

v2.0.0

Tự Chủ

The Autonomy

Autonomous Workflows — watch daemon, YAML workflow engine, pre-commit 2-tier scanning, multi-agent pipelines.

Watch Daemon — agentbox watch — react to file changes, trigger workflows automatically.
Pre-commit 2-Tier Scan — Tier 1 regex (<1s) + Tier 2 heuristic. Block secrets, SQLi, CMDi before commit.
Workflow YAML Engine — Declarative steps, conditions, budget control. Like GitHub Actions for AI agents.

Epics

Features

4,030

Tests

Epics 165–168

2 tháng 4, 2026

v2.1.0

Tiến Hoá

The Evolution

Agent Factory + Self-Update — auto-generate agents from descriptions, self-evolve from feedback, A/B testing.

Agent Factory — Describe in natural language → 6-stage pipeline generates agent with persona + knowledge + policies.
Self-Update — Agents analyze feedback → modify knowledge → A/B test → auto-version bump.
20 Demo Agents — Full catalog: security, quality, testing, performance, docs, style, CI, i18n, changelog.

Epics

Features

4,046

Tests

Epics 169–172

v2.1.0

Tiến Hoá

The Evolution

Agent Factory + Self-Update — auto-generate agents from descriptions, self-evolve from feedback, A/B testing.

Agent Factory — Describe in natural language → 6-stage pipeline generates agent with persona + knowledge + policies.
Self-Update — Agents analyze feedback → modify knowledge → A/B test → auto-version bump.
20 Demo Agents — Full catalog: security, quality, testing, performance, docs, style, CI, i18n, changelog.

Epics

Features

4,046

Tests

Epics 169–172

2 tháng 4, 2026

1 tháng 4, 2026

v1.7.0

Nhận Thức

The Awareness

Context-Aware Execution — agents auto-detect project type, language, framework, and dependencies. Zero config.

5 Language Detectors — Go, Node/TypeScript, Python, Rust, Java — auto-detected from signal files
Monorepo Detection — Turborepo, pnpm, Nx, Lerna — detect root + sub-project context
Token-Budgeted Injection — Max 2000 tokens context injected into agent prompt. Cache <10ms.

Epics

Features

3,869

Tests

Epics 158–159

1 tháng 4, 2026

v1.7.0

Nhận Thức

The Awareness

Context-Aware Execution — agents auto-detect project type, language, framework, and dependencies. Zero config.

5 Language Detectors — Go, Node/TypeScript, Python, Rust, Java — auto-detected from signal files
Monorepo Detection — Turborepo, pnpm, Nx, Lerna — detect root + sub-project context
Token-Budgeted Injection — Max 2000 tokens context injected into agent prompt. Cache <10ms.

Epics

Features

3,869

Tests

Epics 158–159

1 tháng 4, 2026

v1.8.0

Đơn Giản

The Simplicity

CLI UX Redesign — Docker/npx-style agentbox <agent> shorthand, BMAD-style override layer, project-local agents.

agentbox <agent> Shorthand — 1 command to run any agent. Auto-pull, one-shot, pipe, --serve modes.
Override Layer (BMAD-style) — .base/ + overrides/ + knowledge/ — customize agents per-project, team share via git.
.agentbox/config.yaml — npm-like manifest: agents, default, shortcuts, shared_knowledge. agentbox install after clone.

Epics

Features

3,950

Tests

Epics 160–162

v1.8.0

Đơn Giản

The Simplicity

CLI UX Redesign — Docker/npx-style agentbox <agent> shorthand, BMAD-style override layer, project-local agents.

agentbox <agent> Shorthand — 1 command to run any agent. Auto-pull, one-shot, pipe, --serve modes.
Override Layer (BMAD-style) — .base/ + overrides/ + knowledge/ — customize agents per-project, team share via git.
.agentbox/config.yaml — npm-like manifest: agents, default, shortcuts, shared_knowledge. agentbox install after clone.

Epics

Features

3,950

Tests

Epics 160–162

1 tháng 4, 2026

31 tháng 3, 2026

v1.6.0

Kết Nối

The Bridge

Universal MCP Local Server — agents accessible from any IDE, Slack bot, CI/CD, or script via stdio + HTTP dual transport.

MCP Server (stdio + HTTP) — agentbox mcp serve hosts agents as tools for Claude Code, Cursor, VS Code, and any MCP client
Agent-as-Tool Pattern — ask_{name}(question) → persona + knowledge + LLM → response. Hot reload on pull.
Agent Mesh — Agents call agents internally via MCP routing. Multi-agent collaboration.

Epics

Features

3,825

Tests

Epics 155–157

31 tháng 3, 2026

v1.6.0

Kết Nối

The Bridge

Universal MCP Local Server — agents accessible from any IDE, Slack bot, CI/CD, or script via stdio + HTTP dual transport.

MCP Server (stdio + HTTP) — agentbox mcp serve hosts agents as tools for Claude Code, Cursor, VS Code, and any MCP client
Agent-as-Tool Pattern — ask_{name}(question) → persona + knowledge + LLM → response. Hot reload on pull.
Agent Mesh — Agents call agents internally via MCP routing. Multi-agent collaboration.

Epics

Features

3,825

Tests

Epics 155–157

31 tháng 3, 2026

30 tháng 3, 2026

v1.5.0

Sẵn Sàng

The Readiness

CLI Production Hardening — E2E demo, agent package format, MCP tool-calling engine, rate limit retry, Windows support.

CLI E2E Demo — 19-step end-to-end demo: init → pull → run → scan → report. Fully automated.
MCP Tool-Calling Engine — Local tool execution, per-agent MCP config, connected mode with server-side tools/call.
Agent Package Format — .agentbox format: agent.yaml v2 + persona.md + knowledge/ + tools/ + policies/
Production Hardening — Rate limit retry, Windows support, graceful shutdown, NemoClaw mandatory enforcement.

Epics

Features

3,810

Tests

Epics 136–154

v1.5.0

Sẵn Sàng

The Readiness

CLI Production Hardening — E2E demo, agent package format, MCP tool-calling engine, rate limit retry, Windows support.

CLI E2E Demo — 19-step end-to-end demo: init → pull → run → scan → report. Fully automated.
MCP Tool-Calling Engine — Local tool execution, per-agent MCP config, connected mode with server-side tools/call.
Agent Package Format — .agentbox format: agent.yaml v2 + persona.md + knowledge/ + tools/ + policies/
Production Hardening — Rate limit retry, Windows support, graceful shutdown, NemoClaw mandatory enforcement.

Epics

Features

3,810

Tests

Epics 136–154

30 tháng 3, 2026

25 tháng 3, 2026

v1.1.0

Vuốt Sắc

The Claw

NemoClaw mandatory integration — kernel-level agent sandboxing with Landlock + seccomp + netns. L0-L3 permission model, binary integrity verification.

NemoClaw Sandbox Runtime — Kernel-isolated execution with Landlock + seccomp + network namespaces. No fallback — mandatory for all agents.
L0–L3 Permission Model — 4-tier trust: Untrusted → Community → Verified → Enterprise. Kernel-enforced restrictions per tier.
Binary Integrity — SHA-256 hash verification on every NemoClaw startup. Tamper-proof agent isolation.
IPC Protocol — JSONL over stdio pipe between CLI and NemoClaw child process. Cross-platform: native Linux, Docker fallback Win/macOS.

Epics

Features

3,785

Tests

Epics 130–135

25 tháng 3, 2026

v1.1.0

Vuốt Sắc

The Claw

NemoClaw mandatory integration — kernel-level agent sandboxing with Landlock + seccomp + netns. L0-L3 permission model, binary integrity verification.

NemoClaw Sandbox Runtime — Kernel-isolated execution with Landlock + seccomp + network namespaces. No fallback — mandatory for all agents.
L0–L3 Permission Model — 4-tier trust: Untrusted → Community → Verified → Enterprise. Kernel-enforced restrictions per tier.
Binary Integrity — SHA-256 hash verification on every NemoClaw startup. Tamper-proof agent isolation.
IPC Protocol — JSONL over stdio pipe between CLI and NemoClaw child process. Cross-platform: native Linux, Docker fallback Win/macOS.

Epics

Features

3,785

Tests

Epics 130–135

25 tháng 3, 2026

24 tháng 3, 2026

v1.0.0

Khởi Đầu

The Beginning

v1.0 release — observability, production deployment, agent runtime pipeline with Claude Code CLI, BYOK key vault, runtime configuration, and user journey demo flows.

Agent Runtime Pipeline — RuntimeFactory with Docker and Claude Code CLI/SDK runtimes, MCP server config
BYOK Key Vault — Bring Your Own Key — AES-256-GCM encrypted storage for Anthropic, OpenAI, and Google API keys
Runtime Configuration — Per-agent model, budget, timeout, tool permissions, and MCP server settings in Builder Lab
Production Deployment — Docker Compose production stack with Caddy reverse proxy and health checks
Performance & Observability — k6 load tests, Sentry, structured logging, Redis caching, Lighthouse CI

Epics

141

Features

3,248

Tests

Epics 83–129

v1.0.0

Khởi Đầu

The Beginning

v1.0 release — observability, production deployment, agent runtime pipeline with Claude Code CLI, BYOK key vault, runtime configuration, and user journey demo flows.

Agent Runtime Pipeline — RuntimeFactory with Docker and Claude Code CLI/SDK runtimes, MCP server config
BYOK Key Vault — Bring Your Own Key — AES-256-GCM encrypted storage for Anthropic, OpenAI, and Google API keys
Runtime Configuration — Per-agent model, budget, timeout, tool permissions, and MCP server settings in Builder Lab
Production Deployment — Docker Compose production stack with Caddy reverse proxy and health checks
Performance & Observability — k6 load tests, Sentry, structured logging, Redis caching, Lighthouse CI

Epics

141

Features

3,248

Tests

Epics 83–129

24 tháng 3, 2026

17 tháng 3, 2026

v0.9.0

Bình Minh

The Dawn

Enterprise-ready — team workspaces, real-time collaboration, horizontal scaling, Storybook, and code review hardening.

Team Workspaces — Organizations with role-based access, invites, and workspace switching
Real-time Collaboration — Live presence indicators and concurrent editing support
Security Hardening — 41 fixes from adversarial code review (8C/13H/20M/16L findings)
Storybook — Component library documentation with Chromatic visual testing

Epics

Features

2,052

Tests

Epics 61–82

17 tháng 3, 2026

v0.9.0

Bình Minh

The Dawn

Enterprise-ready — team workspaces, real-time collaboration, horizontal scaling, Storybook, and code review hardening.

Team Workspaces — Organizations with role-based access, invites, and workspace switching
Real-time Collaboration — Live presence indicators and concurrent editing support
Security Hardening — 41 fixes from adversarial code review (8C/13H/20M/16L findings)
Storybook — Component library documentation with Chromatic visual testing

Epics

Features

2,052

Tests

Epics 61–82

17 tháng 3, 2026

14 tháng 3, 2026

v0.7.0

Cánh Buồm

The Sails

User ecosystem — tabbed agent detail pages, creator profiles, documentation hub, feedback system, Go CLI, and distribution.

Agent Detail Pages — Tabbed interface: Overview, Reviews, API, Changelog, MCP
Creator Storefronts — Public creator profiles with portfolio and earnings dashboard
Documentation Hub — MDX-powered docs with search, code highlighting, and API reference
Go CLI — Full-featured CLI with Cobra/Viper for agent management

Epics

Features

1,660

Tests

Epics 43–60

v0.7.0

Cánh Buồm

The Sails

User ecosystem — tabbed agent detail pages, creator profiles, documentation hub, feedback system, Go CLI, and distribution.

Agent Detail Pages — Tabbed interface: Overview, Reviews, API, Changelog, MCP
Creator Storefronts — Public creator profiles with portfolio and earnings dashboard
Documentation Hub — MDX-powered docs with search, code highlighting, and API reference
Go CLI — Full-featured CLI with Cobra/Viper for agent management

Epics

Features

1,660

Tests

Epics 43–60

14 tháng 3, 2026

11 tháng 3, 2026

v0.5.0

Nền Tảng

The Foundation

Advanced platform — multi-LLM support, agent templates, versioning, pipeline DAG editor, marketplace economics, and OWASP security hardening.

Multi-LLM Support — Claude, GPT-4, Gemini with model selection and fallback
Agent Templates — Pre-built templates for common use cases with one-click cloning
Pipeline DAG Editor — Visual workflow builder with Action, Service, and Gate nodes
Marketplace Economics — Creator revenue sharing, pricing tiers, and analytics
OWASP Hardening — 54/54 security checks passing with CSRF, XSS, SQLi protection

Epics

Features

1,200

Tests

Epics 20–42

11 tháng 3, 2026

v0.5.0

Nền Tảng

The Foundation

Advanced platform — multi-LLM support, agent templates, versioning, pipeline DAG editor, marketplace economics, and OWASP security hardening.

Multi-LLM Support — Claude, GPT-4, Gemini with model selection and fallback
Agent Templates — Pre-built templates for common use cases with one-click cloning
Pipeline DAG Editor — Visual workflow builder with Action, Service, and Gate nodes
Marketplace Economics — Creator revenue sharing, pricing tiers, and analytics
OWASP Hardening — 54/54 security checks passing with CSRF, XSS, SQLi protection

Epics

Features

1,200

Tests

Epics 20–42

11 tháng 3, 2026

8 tháng 3, 2026

v0.3.0

Rễ Cây

The Roots

Quality infrastructure — E2E test suite, CI/CD pipelines, Shadcn/UI migration, AppShell, dark mode, and accessibility compliance.

E2E Test Suite — Playwright tests with page objects, fixtures, and CI sharding
Shadcn/UI Migration — Complete component library migration with design tokens
Dark Mode — System-aware theme switching with CSS custom properties
Accessibility — WCAG 2.1 AA compliance with axe-core automated checks

Epics

Features

480

Tests

Epics 8–19

v0.3.0

Rễ Cây

The Roots

Quality infrastructure — E2E test suite, CI/CD pipelines, Shadcn/UI migration, AppShell, dark mode, and accessibility compliance.

E2E Test Suite — Playwright tests with page objects, fixtures, and CI sharding
Shadcn/UI Migration — Complete component library migration with design tokens
Dark Mode — System-aware theme switching with CSS custom properties
Accessibility — WCAG 2.1 AA compliance with axe-core automated checks

Epics

Features

480

Tests

Epics 8–19

8 tháng 3, 2026

6 tháng 3, 2026

v0.1.0

Hạt Giống

The Seed

Foundation release — authentication, marketplace browsing, agent builder lab, Docker execution engine, and admin dashboard.

Auth & User System — Supabase Auth with email/password, RLS policies, session management
Agent Marketplace — Browse, search, filter agents with category and tag system
Builder Lab — Three-phase guided workflow: Discovery → Logic → Output
Execution Sandbox — Docker-isolated agent execution with SSE streaming
MCP Integration — Model Context Protocol support for IDE connections

Epics

Features

120

Tests

Epics 1–7

6 tháng 3, 2026

v0.1.0

Hạt Giống

The Seed

Foundation release — authentication, marketplace browsing, agent builder lab, Docker execution engine, and admin dashboard.

Auth & User System — Supabase Auth with email/password, RLS policies, session management
Agent Marketplace — Browse, search, filter agents with category and tag system
Builder Lab — Three-phase guided workflow: Discovery → Logic → Output
Execution Sandbox — Docker-isolated agent execution with SSE streaming
MCP Integration — Model Context Protocol support for IDE connections

Epics

Features

120

Tests

Epics 1–7

6 tháng 3, 2026

Next-Gen Agent Platform

v4.0.0

Vô Hạn

— The Infinite

Multi-modal agents — code + vision + voice in one conversation.

Target: Q3 2026

Agents Built-in

187

Epics Shipped

4,394

Total Tests

Autopilot Presets

CLI Packages

Event Types

Universal Runtime — NemoClawLite

Kernel-level isolation (Landlock + seccomp + netns) for every tier. Same security, different features.

Free

NemoClawLite + Local LLM

Full CLI, kernel isolation, Ollama local LLM, unlimited agents, 20 built-in agents, autopilot

Popular

Pro

$19/month

NemoClawLite + Cloud LLM

Cloud hosting, team features, Claude / GPT integration, memory sync, PR review bot, priority support

Enterprise

NemoClawLite + Custom LLM

Full policy control, audit trail, SSO, compliance, private registry, mission control dashboard, SLA

What's Coming in v2.0

6 major features across ecosystem, intelligence, security, and enterprise

Multi-Modal Agents

IntegrationPlanned

Agents that understand code, screenshots, design mockups, and voice commands. Vision + code + text in one conversation.

Federated Agent Networks

PlatformResearch

Cross-organization agent collaboration. Shared knowledge without sharing code. Privacy-preserving federated learning.

Real-Time Human-Agent Collab

IntelligencePlanned

Live pair programming with agents. Shared cursor, inline suggestions, voice commands. Google Docs for code.

Agent Marketplace v2

EcosystemPlanned

Workflow marketplace: sell curated agent packs + triggers + config as bundles. Recurring revenue model.

Mobile Companion PWA

EnterprisePlanned

Push notifications, approval swipe, daily summary cards. Monitor autopilot on the go.

Self-Hosted Air-Gapped

SecurityPlanned

On-premise deployment for regulated industries. Local model inference, no external calls.

Development Progress

Track every epic across v1.0, v1.1, and the upcoming v2.0 Agent Ecosystem — with real-time progress bars and a visual Gantt chart.

View Gantt Chart & Progress

Beyond v2.0

What's on the horizon

Planned

Mobile Companion App

iOS and Android app for monitoring agent executions and quick interactions on the go

Planned

Self-Hosted On-Premise

Air-gapped deployment option for regulated industries with local model inference

Planned

Marketplace API

Public REST + GraphQL API for third-party integrations and custom agent storefronts

Planned

Multi-Tenant White Label

Resellable platform with custom branding, domains, and tenant-isolated infrastructure