The Public Prompt License (PPL) is the open-source standard for the AI-Native era. Protect your prompts, define your agent's soul, and build the cognitive commons.
Traditional licenses like MIT and AGPL were written for compiled code. They don't understand AI Agents.
The Problem: If you run an agent as a service, AGPL is ambiguous about whether you must share the Python wrapper or the System Prompt. SSPL tries to fix this but claims the entire "Service Stack," which is overreaching.
The Solution: PPL defines "Prompt Source"—the cognitive logic, personas, and tool descriptions—and separates it from the infrastructure.
Read the philosophy: From GPL to PromptsExplicitly defined as the natural language instructions, personas, and schemas.
Explicitly excludes the OS, inference engine, and hardware drivers. No "SSPL Trap".
Choose the variant that fits your project's needs.
The "MIT" Variant
The "Apache" Variant
The "Service" Variant
Start using PPL in your project in 2 steps.
Copy the text of your chosen variant (M, A, or S) into a file named LICENSE at the root of your repo.
Add a copyright header to your main prompt files (e.g., system.md).
<!--
Copyright (c) 2025 Your Name
Licensed under the Public Prompt License - MIT Variant (PPL-M)
See LICENSE for details.
-->No. PPL-S explicitly excludes "Underlying Software" (inference engines, OS, network stack) from the disclosure requirement. You only need to share the Prompt Source (the cognitive logic).
Never share secrets. The definition of Prompt Source excludes "User Session Data" and implies that configuration secrets should be redacted. Share the structure, not your credentials.
Not yet. PPL is currently a Draft (v0.1). We have designed it to align with the Open Source Definition (OSD) and plan to submit it for review after gathering community feedback.