An open-source specification for democratic political participation.
🌐 open-politos.org · 📦 github.com/open-politos
Political parties today are often opaque, slow to respond, and easily captured by organized minorities or powerful individuals. Manifestos are written in back rooms. Promises are broken without accountability. Citizens have little insight into how decisions are actually made.
PolitOS is a different approach.
It is an open specification that describes how anyone — a group of citizens, a movement, a new party — can build a political organization where the rules are written in public, decisions are made transparently, and an AI assistant helps translate collective will into concrete policy positions.
No black boxes. No back rooms. Everything auditable by anyone.
PolitOS is currently a specification, not a running application. This repository defines the rules, processes, and agent architectures for a transparent political organization. There is no code to install or run yet.
We are designing the blueprint first — openly and collaboratively — before building software on top of it. This is intentional: the rules matter more than the code.
Contributions to the specification are welcome. See Get Involved.
Imagine a political party where:
- The party's positions are not decided by a powerful leader, but generated by an AI that the members collectively govern
- Every citizen can see exactly what the AI has been told to believe and why
- Changing the AI's position on any issue requires a structured, transparent democratic process — not a phone call between insiders
- The party has an AI spokesperson that can only say what has been officially decided — never improvising, never spinning
PolitOS describes the infrastructure to build exactly this. It is configurable, open-source, and free for anyone to use.
PolitOS is not a political party. It has no ideology. It does not tell you what to believe.
It is a specification — like a legal framework or a voting system — that any group can use to organize themselves more transparently and democratically. The values and positions of a party built on PolitOS are entirely determined by its members.
1. Your organization's rules live in a public document. Everything — how members join, how decisions are made, what the AI is allowed to say, what it can never say — is written down and publicly readable. No hidden bylaws.
2. Members govern the AI together. The AI does not set its own agenda. Members propose changes to what it knows and how it reasons. Every proposal goes through a structured deliberation process before it can be adopted.
3. The AI speaks for the party — but only what the party has decided. A dedicated AI agent can act as spokesperson or party representative. It communicates clearly and consistently, but it is constitutionally incapable of going off-script. Every statement it makes is logged and traceable.
4. Safeguards prevent manipulation. Organized factions cannot simply outvote everyone else. The system uses random selection of decision panels, mandatory waiting periods, and transparency requirements to make manipulation expensive and visible.
5. Anyone can verify everything. Every decision, every change, every AI output is logged in a public audit trail. Journalists, citizens, and members can always see what happened and when.
Citizens and activists who want to build a political movement with genuine transparency at its core.
Existing parties looking to modernize their internal decision-making and member participation.
Researchers and democracy practitioners exploring new models of AI-augmented governance.
Developers who want to contribute to open democratic infrastructure.
You do not need to be a developer to contribute. Right now, we are shaping the specification and need:
- People who understand political organizing and party law
- Designers who can think about how governance tools should work
- Writers who can translate complex governance concepts for the public
- Developers who want to help design democratic infrastructure
→ Website → Start a discussion → Read the contribution guide
This repository contains the complete specification for a PolitOS-based organization:
PolitOS/
├── constitution/ # Hardcoded ethical & legal constraints
├── knowledge-base/ # Versioned knowledge base fed to the policy agent
├── prompts/ # Governed prompt configs per policy domain
├── governance/ # Voting rules, deliberation protocols, sortition logic
├── membership/ # Membership tiers, onboarding flows, rights matrix
├── agents/
│ ├── policy-engine/ # Core policy generation agent
│ ├── representative/ # Public spokesperson & party leader persona
│ │ ├── persona.yaml # Name, voice, communication style
│ │ ├── boundaries.yaml # What the agent can/cannot say publicly
│ │ └── escalation.yaml # When to trigger new governance proposals
│ ├── advocate/ # Challenges every proposal with counter-arguments
│ ├── compliance/ # Validates outputs against constitution layer
│ ├── summarizer/ # Produces citizen-facing summaries of decisions
│ ├── kb-curator/ # Proposes KB updates based on new information
│ └── moderator/ # Manages deliberation phases
├── adapters/
│ └── base/ # Abstract LLM interface (model-agnostic)
├── audit-log/ # Append-only log of all governance decisions
├── config/
│ └── party.config.yaml # Organization-specific configuration
└── docs/
PolitOS is designed to be model-agnostic. The specification defines an abstract adapter interface that can be implemented for any LLM backend:
llm:
provider: "claude" # claude | openai | ollama | mistral | custom
model: "claude-opus-4-6"
fallback: "ollama/llama3"MIT — fork freely, govern responsibly.
"A party is not its leaders. It is its rules."