Public Prompt License (PPL)

Status: Draft 0.2 (Request for Comment)

Read the announcement: Shipping PPL: A License for the Age of Promptware

The Public Prompt License (PPL) is a family of open-source licenses designed specifically for the "AI-Native" era, where the core logic of an application lives in natural language prompts rather than compiled code.

The Problem: "English is the new Source Code"

Traditional open-source licenses (MIT, Apache, GPL) were written for software: compiled binaries, source code files, and linked libraries. They do not cleanly map to AI Agents, where the "source code" is a set of system prompts, persona definitions, and few-shot examples.

• MIT/Apache work for text, but don't address the specific needs of "Agent Services" (running an agent as an API).
• AGPL triggers on "remote interaction," but its definition of "Source Code" is ambiguous when applied to LLM weights vs. prompts vs. Python wrappers.
• SSPL extends copyleft to the entire "Service Stack" (OS, backups, hardware drivers). PPL-S takes a narrower approach to align with OSI standards, limiting the scope strictly to the agent's cognitive logic.

The Solution: PPL

PPL v0.2 takes a "Wrapper" approach. Instead of writing new custom licenses, PPL simply applies standard Open Source licenses (MIT, Apache, AGPL) and extends their scope to explicitly include "Prompt Source".

It separates the cognitive logic (the prompts) from the infrastructure (the inference engine).

The Three Variants

License	Analog	Best For	Key Feature
PPL-M	MIT	Maximal Adoption	It IS the MIT License, extended to cover prompts.
PPL-A	Apache 2.0	Corporate Use	It IS the Apache 2.0 License, extended to cover prompts.
PPL-S	AGPL v3	Agent Services	It IS the AGPL v3 License, extended to cover prompts.

Definitions

All PPL licenses rely on a shared set of definitions found in DEFINITIONS-0.2.md. This ensures that "Prompt Source" is consistently defined as the narrative and cognitive core of the agent, explicitly excluding the underlying inference engine (vLLM, Ollama, etc.).

How to Apply

1. Choose your variant

• Use PPL-M if you want your prompts to be used everywhere, even in closed-source products.
• Use PPL-A if you want patent protection and structured attribution.
• Use PPL-S if you believe in "Prompt Copyleft": if someone profits from running your agent as a service, they should contribute their improvements back to the community.

2. Add the License File

Copy the text of your chosen license from the LICENSES/ directory to a file named LICENSE in your repository.

3. Add the Header

Add a header to your main prompt files (e.g., system.md, agent.yaml):

Copyright (c) [Year] [Your Name]
Licensed under the Public Prompt License - [Variant] (PPL-[Variant])
See LICENSE for details.

Roadmap to OSI

This project is structured to prepare for future submission to the Open Source Initiative (OSI).

• Phase 1 (Now): Draft 0.2 release and community feedback.
• Phase 2: Adoption by pilot projects (e.g., Promptware OS).
• Phase 3: Legal review and refinement.
• Phase 4: Formal OSI submission.

Authors

• Felix - AI Co-founder (Founder of Ship.Fail)
• Huan Li - Human Co-founder & Architect

Contributing

We welcome feedback from developers, lawyers, and AI researchers. Please open an issue to discuss definitions, scope, or wording.