Build multi‑agent AI systems
with the world's first Software 3.0 stack

3-minute Introduction to Playbooks AI

Open Standard

Playbooks Language

Plain English, with a hint of markdown. No complex syntax to learn.
If you can write instructions for a human, you can write a Playbooks program.

customer_support.pb

Open Source (MIT License)

Playbooks Runtime

Execute Playbooks programs to get multi-agent systems.

pip install playbooks

Unified Call Stack

English and Python playbooks can call each other on the same stack and share program state seamlessly.

Natively Multi-Agent

Built-in support for multi-agent interactions like having multi-party meetings and collaborative sessions.

Memory Layers

Scratchpad for short-term memory, artifacts for long-term memory, and stack-based automated LLM context management.

Semantic Triggers

Event-driven execution based on semantic conditions and intelligent context switching.

Open Standard

Playbooks Assembly Language

A standard that defines how to annotate natural language programs for reliable execution on LLMs. A semantic compiler transforms Playbooks programs into Playbooks Assembly Language.

.pb

Playbooks Compiler

.pbasm

customer_support.pbasm

Free to install and use

Playbooks Debugger

VSCode extension to step-debug through your Playbooks programs like traditional code. Use breakpoints, inspect variables and navigate call stack.

Finally, you can step-debug through LLM prompts!

customer_support.pbasm

Early Adopters

Building with Playbooks

Forward-thinking companies are exploring Playbooks for their AI agent systems

ODYSTRA is using Playbooks to build the core intelligence fabric for high-value, high-trust knowledge work across business domains - transforming static enterprise information into dynamic, collaborative, and autonomous processes.

odystra.com →

Commercial License

PlaybooksLM

For Enterprise deployments, instead of using third party frontier models like Claude, use PlaybooksLM hosted within your infrastructure or hosted on Playbooks Cloud (coming soon).

10x

Faster Execution

Token throughput vs Claude Sonnet 4.0

10x

Lower Costs

Inference costs vs Claude Sonnet 4.0

Higher Reliability

For Playbooks execution

On-Premise

Complete deployment control

Commercial License

Enterprise

Enterprise-grade multi-agent system with compliance, security, scalability, and observability.

Security & Scalability

On-premise deployment and SLM-based execution

Full Observability

LangFuse compatible, audit trails, and execution tracing

Framework Integration

Works with LangGraph, Google ADK, Autogen, and MCP

Unlock Enterprise Capabilities

Open Source

Context Engineering

Playbooks runtime offers advanced context engineering capabilities to optimize LLM interactions and manage complex workflows efficiently.

Stack-based Context Management

Automatic context management that follows your program's call stack, ensuring optimal context for each execution level.

Incremental Compaction

Built-in context compaction without extra LLM calls, keeping essential information while reducing token usage.

LLM Caching Optimization

Built-in optimization for LLM caching to reduce costs and improve performance through intelligent context reuse.

Context Augmentation

Easy-to-use techniques for adding file content, retrieval results, or outputs from other playbooks directly into context.

Welcome to Software 3.0!

Let's build something magicalwith Playbooks AI

Amol Kelkar (Founder, Playbooks AI)

Get Started Now

Getting Started

Documentation Getting Started Tutorial

Links

GitHub Repository Blog