Skip to content
Alessio Rocchi edited this page Jan 29, 2026 · 3 revisions

Welcome to aistack Wiki

Production-grade multi-agent orchestration for Claude Code with adversarial validation, persistent memory, and real-time web dashboard.

npm version CI codecov


Quick Links

Getting Started

Core Concepts

Agent Guides

MCP Tools Reference

Recipes & Examples

Advanced Topics

Reference


What is aistack?

aistack is a production-grade multi-agent orchestration framework that enables you to:

  • Spawn specialized AI agents - 11 agent types with distinct capabilities
  • Coordinate complex tasks - Hierarchical task management with message bus
  • Maintain persistent memory - SQLite with FTS5 full-text and vector search
  • Integrate with Claude Code - 46 MCP tools for seamless IDE integration
  • Run adversarial validation - Coder ↔ Adversarial review loops
  • Monitor in real-time - React web dashboard with live updates

System at a Glance

11 agents · 46 MCP tools · 6 LLM providers · SQLite + FTS5 · Web dashboard · Adversarial review

Architecture

Claude Code ←→ MCP Server ←→ Agent Manager ←→ 11 Agent Types
                    ↓              ↓
              Memory Manager   Task Queue
                    ↓              ↓
               SQLite + FTS5  Message Bus

Key Features

  • Multi-Agent Orchestration: Coordinate coder, researcher, tester, reviewer, adversarial, architect, coordinator, analyst, devops, documentation, and security-auditor agents
  • Adversarial Review Loop: Iterative code validation with coder ↔ adversarial feedback (max 3 iterations)
  • Persistent Memory: SQLite with FTS5 full-text search and optional vector embeddings
  • MCP Integration: 46 tools exposed via Model Context Protocol for Claude Code
  • 6 LLM Providers: Anthropic, OpenAI, Ollama, ClaudeCode CLI, Gemini CLI, Codex
  • Web Dashboard: React 18 + Material-UI with real-time WebSocket updates
  • GitHub Integration: Issues, PRs, repo info via gh CLI
  • Plugin System: Runtime extensibility without modifying core
  • Drift Detection: Semantic drift monitoring with configurable thresholds to prevent task scope creep
  • Consensus Checkpoints: Human-in-the-loop approval for high-risk tasks
  • Agent Identity: Persistent agent identities with lifecycle management and audit trails
  • Resource Monitoring: Track and limit agent resource consumption with automatic intervention

Quick Start

# Install
npm install @blackms/aistack

# Initialize
npx @blackms/aistack init

# Add to Claude Code
claude mcp add aistack -- npx @blackms/aistack mcp start

# Start web dashboard
npx @blackms/aistack web start

See Installation for detailed setup instructions.


Popular Topics

For New Users

  1. Quick Start Tutorial - Get up and running in 5 minutes
  2. Understanding Agents - Learn about agent types
  3. Memory System - How persistent memory works

For Developers

  1. Agent Overview - Detailed agent capabilities
  2. Tools Overview - All 46 MCP tools
  3. Programmatic API - TypeScript SDK

For Advanced Users

  1. Plugin Development - Extend aistack
  2. Custom Agent Types - Create custom agents
  3. Workflow Engine - Multi-phase workflows

Community & Support


Contributing

Contributions are welcome! See the Contributing Guide for details.


Last updated: 2026-01-29 | Version: 1.5.3

Clone this wiki locally