Agent 0+

A personal AI assistant framework — forked from Agent Zero, enhanced with 30+ tools, self-healing intelligence, and a technician problem-solving mindset.

---

What is Agent 0+?

Agent 0+ is a heavily enhanced fork of Agent Zero — an open-source, fully autonomous AI agent framework. While Agent Zero provides the core agentic loop, Agent 0+ adds:

• 30+ custom tools for real-world tasks (email, GitHub, weather, video analysis, image generation, and more)
• Self-healing technician mindset — the agent diagnoses errors, installs missing dependencies, and searches the web for solutions
• Plugin system for messaging channels (Discord, Slack, WhatsApp, Telegram, Matrix)
• Enhanced browser backend with multiple provider support
• Skill-based architecture using the open SKILL.md standard

Agent 0+ thinks like a technician. When something fails, it reads the error, diagnoses the root cause, installs what's needed, tries alternative approaches, and verifies the fix — all autonomously.

---

⚡ Quick Start

# Clone the repo
git clone https://github.com/David2024patton/agent-zero.git
cd agent-zero

# Copy environment template
cp .env.example .env
# Edit .env with your API keys

# Pull and run with Docker
docker compose up -d

# Visit http://localhost:50001 to start

Requirements

• Docker & Docker Compose
• At least one LLM API key (OpenAI, Google Gemini, Z.AI, Mistral, or any LiteLLM-supported provider)

---

🧰 30+ Custom Tools

Agent 0+ ships with tools that go far beyond the Agent Zero defaults:

Communication & Messaging

Tool	Description
email_tool	Send/receive emails via SMTP/IMAP (configurable hosts)
message_channel	Unified messaging across Discord, Slack, Telegram, WhatsApp, Matrix

Developer & DevOps

Tool	Description
github_tool	Repos, issues, PRs, file contents, gists, workflow listing
hostinger_tool	DNS records, SSL status, hosting management
browser_agent	Full browser automation with multiple backend support
browser_read	Read and extract content from web pages
document_to_markdown	Convert documents to markdown format

AI & LLM Integration

Tool	Description
gemini_tool	Google Gemini API (generate, analyze, count tokens)
mistral_tool	Mistral AI API integration
zai_tool	Z.AI API integration
huggingface_tool	Model inference, search models/datasets
image_gen_tool	DALL-E image generation, editing, and variations
whisper_tool	OpenAI Whisper speech-to-text transcription
summarize_tool	URL/text summarization (bullet, paragraph, academic styles)
swarm	Multi-agent parallel analysis with tiered model selection

Productivity

Tool	Description
notion_tool	Pages, databases, blocks — full Notion API
trello_tool	Boards, cards, labels, search

Media & Data

Tool	Description
weather_tool	Current weather & forecasts (metric/imperial, feels-like temp)
video_transcript_tool	YouTube transcripts, subtitles, keyword search
video_frames_tool	Extract frames, thumbnails, GIF conversion, resize

Skills (Auto-loaded)

Skill	Description
system-report	Generate PDF system health reports with CPU/RAM/disk charts
search-engine	Web search via SearXNG
weather	Weather lookups via wttr.in and Open-Meteo
browser-agent	Browser automation guidance
And 25+ more...	Email, GitHub, Notion, Trello, scheduler, etc.

---

🧠 Self-Healing Technician Mindset

This is the core enhancement that sets Agent 0+ apart. The agent is instructed to think like a real technician:

When something fails:

1. Read the error — don't blindly retry
2. Diagnose the cause — missing package? bad config? wrong path?
3. Fix it — install deps, set env vars, try alternatives
4. Verify the fix — confirm it actually worked
5. Move on — treat every error as a puzzle to solve

When it doesn't know something:

• Uses search_engine to Google the error message, library docs, or technique
• Reads Stack Overflow, GitHub issues, docs — just like a real technician
• Never guesses blindly — always researches before attempting unfamiliar fixes

Self-repair rules:

• Never repeats the same failing action more than twice without changing approach
• If approach A fails, tries approach B (different library, method, or fallback)
• If both fail, searches the web for how others solved the same problem
• Auto-installs missing packages (pip install, apt-get install, npm install)

---

🔌 Plugin System

Agent 0+ includes a plugin architecture for messaging channels:

Plugin	Description
discord_channel	Discord bot integration
slack_channel	Slack workspace integration
whatsapp_channel	WhatsApp messaging
telegram_channel	Telegram bot integration
matrix_channel	Matrix/Element integration

Plugins are loaded automatically from the plugins/ directory.

---

🛡️ Security

• Secrets management — API keys are stored encrypted and injected at runtime; the agent never sees raw credentials
• Docker isolation — all agent code runs inside a sandboxed Kali Linux container
• .gitignore protection — .env, usr/, and tmp/ directories are excluded from git
• No credentials in code — all tools read secrets from environment variables

Setting up your keys:

1. Copy .env.example to .env
2. Add your API keys to .env
3. Or use the Web UI: Settings → External Services → Secrets

Required keys depend on which tools you want to use:

Key	Tools
OPENAI_API_KEY	image_gen, whisper, summarize
GEMINI_API_KEY	gemini_tool
MISTRAL_API_KEY	mistral_tool
GITHUB_PERSONAL_ACCESS_TOKEN	github_tool
HUGGINGFACE_TOKEN	huggingface_tool
NOTION_API_KEY	notion_tool
TRELLO_API_KEY + TRELLO_TOKEN	trello_tool
HOSTINGER_API_TOKEN	hostinger_tool

---

📁 Project Structure

agent-zero/
├── prompts/                    # All system prompts (fully customizable)
│   ├── agent.system.main.*.md  # Core agent behavior + technician mindset
│   └── agent.system.tool.*.md  # Tool-specific prompts (30+)
├── python/
│   ├── tools/                  # Custom tool implementations
│   ├── helpers/                # Utilities (browser backend, TTS, plugins, etc.)
│   ├── api/                    # API endpoints
│   └── extensions/             # Extension hooks (pre/post LLM call)
├── plugins/                    # Messaging channel plugins
├── skills/                     # Built-in skills
├── usr/skills/                 # User skills (gitignored)
├── webui/                      # Web UI (HTML/CSS/JS)
├── conf/                       # Model providers config
└── docker-compose.yml          # Docker deployment

---

⚙️ Configuration

Model Providers

Agent 0+ supports all LiteLLM-compatible providers:

• OpenAI, Google Gemini, Mistral, Z.AI, Anthropic
• Ollama (local), OpenRouter, CometAPI
• AWS Bedrock, Azure OpenAI
• And many more via LiteLLM

Configure in the Web UI under Settings → Agent Settings or via environment variables with the A0_SET_ prefix.

Browser Backend

Multiple browser automation backends are supported:

• Default Docker-based browser
• External browser connections
• Custom CDP endpoints

---

🔧 Customization

Everything is customizable:

• Prompts — Edit any file in prompts/ to change agent behavior
• Tools — Add new tools in python/tools/ with matching prompts
• Skills — Create SKILL.md files in usr/skills/ for new capabilities
• Plugins — Add messaging channels in plugins/
• Extensions — Hook into the agent loop via python/extensions/

---

🙏 Credits

• Based on Agent Zero by frdel
• Enhanced by David Patton
• Uses the SKILL.md standard developed by Anthropic
• Powered by LiteLLM for multi-provider LLM support

false