Shazam
Autonomous AI Agent Teams
Coordinate multiple Claude AI agents in hierarchical teams. Define roles, delegate tasks, and let your AI company ship software autonomously.
Install
One command. Checks dependencies, builds, and installs.
curl -fsSL https://raw.githubusercontent.com/raphaelbarbosaqwerty/shazam-cli/main/setup.sh | bashCore Capabilities
Everything you need to orchestrate autonomous AI agent teams at scale.
Autonomous Company Model
Hierarchical agent teams with CEO, PM, Dev, QA, and Analyst roles working together under a chain of command.
RalphLoop Execution
Per-company polling loop that picks, executes, and monitors tasks with automatic retry and exponential backoff.
Session Pooling
Reuse Claude sessions across tasks to save tokens and preserve context, recycling after 8 tasks or 15 minutes.
TaskBoard
ETS-backed task management with atomic checkout preventing race conditions. Full lifecycle from pending to completed.
Module Locking
Prevents concurrent edits to the same code module by different agent hierarchies, ensuring safe parallel work.
Subtask Delegation
PMs output JSON subtask blocks that are automatically parsed and assigned to the right agents in the team.
Human-in-the-Loop
PM-created subtasks go through an approval queue before execution, keeping humans in control of critical decisions.
Dual Memory Systems
MemoryBank for per-agent markdown files and SkillMemory for structured knowledge graphs with YAML frontmatter tags.
Real-time Events
EventBus pub/sub with WebSocket streaming for live monitoring via the Flutter dashboard or custom clients.
How It Works
Get from zero to an autonomous AI team in minutes.
Define Your Company
Create a shazam.yaml file describing your agent hierarchy — roles, models, tools, and reporting structure.
Start the Orchestrator
Run shazam start to boot the OTP supervision tree, TaskBoard, SessionPool, and RalphLoop for each company.
Add Tasks
Use the CLI or REST API to submit tasks. The TaskScheduler assigns them to the right agent based on role and availability.
Agents Execute Autonomously
Agents pick up tasks, execute them via Claude Code sessions, delegate subtasks, and report results — all automatically.
Project History
The journey from idea to open-source orchestrator.
The Idea
Born from the need to coordinate multiple AI agents working on the same codebase without conflicts or duplicated effort.
Clawster Era
Initial prototype built in Elixir/OTP, proving that GenServer-based agent management could handle concurrent AI workflows.
Shazam Reborn
Evolved into a full orchestrator with hierarchical companies, session pooling, dual memory systems, and human-in-the-loop approval.
Open Source
Released to the community. Now anyone can build autonomous AI teams with 10 pre-configured agent presets and a powerful CLI.
Ready to Build Your AI Team?
Get started with Shazam in minutes. Define your agents, set up your company, and let them work.