Agent Zero Feature Tour

Explore everything Agent Zero can do — from its web UI to multi-agent cooperation, persistent memory, plugins, MCP, and more.

Web Interface

Beautiful Browser-Based UI

No command line required. Agent Zero runs a full web interface at localhost:50001 where you can chat, monitor agents, view memory, and manage settings.

  • 💬
    Chat InterfaceSend tasks and get responses in a clean, real-time chat UI
  • 🔍
    Agent MonitoringWatch agents work in real-time — see every tool call and result
  • ⚙️
    Settings PanelConfigure API keys, models, memory, and agent behavior
  • 📁
    File BrowserBrowse and manage files in the agent's Docker environment
localhost:50001
You → Analyze the CSV in /data and make a chart
Agent 0 🤖
├─ reading /data/sales.csv
├─ running: python3 analyze.py
├─ creating chart.png with matplotlib
└─ Done ✓
Agent → Created chart.png in /data/
Revenue peaked in Q3 at $47k 📈
Multi-Agent System

Agents That Spawn Agents

Agent Zero can create sub-agents to tackle complex tasks in parallel. Like managing a team — the main agent delegates, monitors, and synthesizes results.

  • 👥
    Dynamic SpawningMain agent creates specialists for research, coding, writing, and more
  • 🔗
    A2A CommunicationAgents pass messages and results between each other seamlessly
  • ⚡
    Parallel ExecutionMultiple agents work simultaneously on different parts of a task
  • 📊
    Result AggregationMain agent synthesizes all outputs into a final coherent answer
Multi-Agent Task
You → Write a market research report on EVs
Agent 0 → spawning 3 agents...
Agent 1 → research latest EV sales data
Agent 2 → analyze competitor landscape
Agent 3 → gather consumer sentiment
... parallel execution ...
Agent 0 → synthesizing results...
→ Report saved to /output/ev_report.pdf
Persistent Memory

An Agent That Learns

Agent Zero's vector memory grows with every interaction. It remembers context, preferences, past tasks, and lessons learned — across sessions.

  • 🧠
    Vector MemorySemantic memory powered by embeddings — finds relevant context automatically
  • 📝
    Solution MemoryRemembers how it solved problems — reuses successful approaches
  • 👤
    User PreferencesLearns your coding style, preferences, and common tasks over time
  • 🔄
    Cross-Session PersistenceMemory persists across restarts when using volume mounts
Memory System
# Week 1
You → I prefer Python with type hints
Agent → Noted ✓ saved to memory
# Week 3 (new session)
You → write a function to parse JSON
Agent → remembered: you use Python + type hints
def parse_json(data: str) -> dict:
return json.loads(data)
Skills & Projects

Reusable Skills in SKILL.md

Define custom skills that Agent Zero can load on demand. Skills are plain Markdown files describing how to perform a task — the agent reads and follows them.

# Example: skills/deploy.md
## Deploy to Production
1. Run tests: npm test
2. Build: npm run build
3. Deploy: railway up
4. Verify health at /api/health
Always check logs after deploy.
  • 📄
    Plain MarkdownWrite skills in simple Markdown — no code required
  • 🔁
    ReusableSave once, use in any task that needs that skill
  • 🏗️
    ProjectsGroup related context, files, and skills into projects
Skills System
# Agent discovers and uses skills
You → deploy the app to production
Agent: loading skill deploy.md...
→ running npm test
✓ All 47 tests passed
→ running npm run build
✓ Build complete
→ running railway up
✓ Deployed to production
→ checking /api/health
✓ Health OK — all done!
MCP — Model Context Protocol

Connect to Any Tool or Service

Agent Zero supports MCP servers — a standard protocol for connecting AI models to external tools, databases, APIs, and services. Hundreds of pre-built integrations available.

Database

PostgreSQL, SQLite

Query databases, run migrations, analyze data

Dev Tools

GitHub, Jira, Linear

Manage issues, PRs, and project boards

Files

Google Drive, S3

Read, write, and organize cloud files

APIs

REST, GraphQL

Connect to any HTTP API or web service

MCP Integration
# MCP server config
{
"mcpServers": {
"github": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-github"]
}
}
}
You → list my open GitHub issues
Agent → using MCP: github.list_issues
Found 7 open issues in your repos
Plugin System

Extend Agent Zero with Plugins

Add new capabilities to Agent Zero without touching the core framework. Plugins are Python modules that add new tools, behaviors, and integrations.

Built-in

🐚 Shell Execution

Run any bash/PowerShell command in the Docker container. Full terminal access.

Built-in

🌐 Web Browser

Browse the web, scrape pages, fill forms, and extract content from URLs.

Built-in

🐍 Python Runtime

Write and execute Python code. Install packages with pip on the fly.

Community

📧 Email Sender

Send emails, read inboxes, manage attachments via Gmail or SMTP.

Community

📅 Calendar

Create events, check availability, schedule meetings on Google Calendar.

Custom

🔧 Build Your Own

Write Python plugins to connect any API or tool your workflow needs.

A2A — Agent-to-Agent

Connect Multiple Agent Zero Instances

A2A (Agent-to-Agent) protocol lets separate Agent Zero instances communicate and delegate tasks to each other over the network.

🏢 Team Setup

Each team member runs their own Agent Zero. Agents collaborate on shared projects automatically.

🌍 Distributed Tasks

Run resource-intensive tasks on a powerful remote machine via A2A delegation.

🤖 Specialized Agents

Create specialist agents (researcher, coder, writer) that the main agent can call on demand.

📖 Start Setup Guide → 🛠 Book Pro Setup — $275