Murray Bookchin argued that the municipality is the scale at which face-to-face democracy remains possible. His program, libertarian municipalism, proposes replacing the nation-state with a confederation of self-governing municipalities organized through popular assemblies. In his later work, Bookchin articulated communalism as a third position: unlike anarchism, it accepts democratic institutions and majority rule; unlike Marxism, it remains fundamentally anti-state and decentralist. PRAXIS encodes Bookchin's architecture directly through the Assembly Playbook, confederal federation, and cooperative economic tools.

In the autonomous regions of North and East Syria, the Kurdish freedom movement adopted Bookchin's vision and built a functioning multi-ethnic democratic governance system serving approximately five million people, constructed during an active civil war. The system operates through communes of 100-350 households, with co-chairs at every level (one of whom must be a woman), parallel women's governance structures, and cooperative economics where profits are split between reinvestment, workers, and community needs.

Since 1994, the Zapatistas have maintained autonomous governance in Chiapas, building their own schools, clinics, courts, and cooperatives entirely outside the Mexican state. Their most radical contribution: they do not seek to seize state power. All authority positions are unpaid and rotating. Communities retain the right to revoke mandates at any time. They have restructured their governance at least three times, most recently in 2023, demonstrating a willingness to adapt structures to changing conditions.

From Bookchin: the neighborhood assembly as the fundamental unit of governance; confederalism as the scaling mechanism; majority-rule democracy over consensus paralysis; the municipality as the arena of transformation. From Rojava: the commune model proven at the scale of millions; gender parity as structural architecture; cooperative economics under democratic control; the lesson that international federation is a survival requirement, not an aspiration. From the Zapatistas: thirty years of proof that autonomous governance outside the state is sustainable; mandar obedeciendo as a design principle for accountability; restorative justice; the willingness to restructure when structures fail; the principle that revolution is a question, not an answer. From Porto Alegre: empirical evidence that participatory budgeting produces superior outcomes at city scale. From Occupy and Minneapolis: the negative proof that protest without infrastructure, consensus without structure, and energy without institutions produces nothing durable. PRAXIS is the technological infrastructure that these traditions required and did not have.

The democratic core. Participatory budgeting, citizen assemblies, formal proposals, structured debate, and voting. The presence of real decisions and real outcomes is what distinguishes PRAXIS from every civic engagement initiative that preceded it.

The one thing PRAXIS builds from scratch. The Comrade Agent tracks your activities, assigns tasks based on the network, and generates your Revolutionary Calendar. Twelve civic roles provide identity within the movement before a participant has completed any traditional organizing work.

Tools that make democratic governance accessible to first-time participants: budget visualization, participation analytics, outcome tracking, and geographic mapping of community needs and power structures.

Under normal conditions, assemblies coordinate through standard tools. When internet infrastructure fails, the practice degrades gracefully through tools designed for exactly that scenario.

How governance decisions translate into economic alternatives that participants can live inside, not merely vote on. Connecting economic decisions to the governance process so they flow from democratic deliberation.

Bookchin's confederalism, implemented as a coordination practice: autonomous assemblies in different cities sharing proposals, outcomes, and methodology without any center having authority over the others. Federation is a practice first and a protocol second. In the near term: shared documentation, regular coordination calls, and a common Assembly Playbook. As the network grows, the Civic Engine connects across cities so that earned standing carries across the federation.

The Fediverse provides decentralized video hosting, image sharing, blogging, music distribution, and discussion, all interconnected through open protocols, all free from corporate algorithms.

If the ecosystem provides the tools, what does PRAXIS itself build? Three things. First, the Civic Engine: the engagement and progression layer that wraps civic participation in quests, roles, and Civic Points. No existing tool does this. Second, the Assembly Playbook: the methodology, facilitation guides, and governance processes that tell a new city how to run their first assembly. Third, the integration glue: the lightweight connectors, deployment guides, and recommended configurations that compose these independent tools into a coherent experience.

Overthrow is a cooperative simulation game built in Godot that models organizing dynamics: crowd behavior, resource allocation, coalition formation. It serves as a recruitment tool through which new participants develop intuition for democratic governance before attending their first assembly, a narrative and lore layer for the Civic Engine's world-building, and content for streaming audiences. Overthrow is a separate project that feeds into PRAXIS, not a dependency of it.

This is a real-world ARG. There are no dice, no ability scores, no hit points. The "game actions" are actual organizing actions. Progression is earned through real experience in the field. Every class corresponds to a function that democratic movements need. Every class has a core mechanic that drives its play pattern.

Classes are not silos. The most effective civic actions emerge from deliberate combinations. First Responder + Field Coordinator: community presence backed by disciplined coordination. Medic + Organizer Ascetic: sustainable long-haul organizing with care infrastructure that prevents burnout. Scout + Saboteur: research that leads to accountability. Culture Jammer + Artist: narrative disruption backed by cultural production that outlasts the news cycle. Eco-Warrior Gardener + Vanguard: land stewardship grounded in unshakable principle. Absurdist + Techie: institutional systems documented and creatively exposed. These are not hypothetical pairings. They are the combinations that have produced results in every movement that understood that a diversity of approaches requires a diversity of roles.

A full organizing team should have at minimum: one presence (First Responder, Field Coordinator, or Vanguard), one support (Medic or Organizer Ascetic), one research (Scout or Techie), one accountability (Saboteur or Absurdist), one narrative (Culture Jammer or Artist), and one wildcard (Eco-Warrior Gardener or any class the local context demands). Not every team will have six people. Not every team will be balanced. But a network that is missing an entire function will eventually feel the gap. The class system makes the gap visible before it becomes a crisis.

Civic Points are not a reward for showing up. They are a formal recognition system for work that materially advances PRAXIS's democratic objectives. The Comrade Agent uses CP values to prioritize suggestions: higher-value actions surface first when the network needs them. The agent tracks your CP history and uses it to calibrate difficulty: a participant who has earned 5,000 CP gets different suggestions than one who has earned 500. CP recognition across the network means that earned standing carries wherever you go. This is the digital implementation of the Zapatista cargo system: rotating, accountable service recognized by the community.

Every participant runs a Comrade Agent: a lightweight AI that knows your class, your skills, your availability, and your ongoing commitments. It does three things. First, it tracks your long-term activities: community garden shifts (week 12 of 52), mutual aid rotations, assembly attendance streaks, cooperative membership tasks, trees planted toward your goal. Your revolutionary history, persistent across sessions. Second, it assigns tasks based on the network: it reads an anonymized feed of what other agents' participants are doing and identifies gaps you can fill. The garden team needs help weeding Thursday and you are an Eco-Warrior with Thursday availability. A Medic is running a care circle and your Scout research would complement it. Three people are planting trees this month and you can join them. Third, it generates your Revolutionary Calendar: a weekly view of committed activities, suggested tasks from the network, overdue items, and progress toward your goals.

It shows up on Monday with a plan. "This week: Wednesday garden shift, Saturday assembly, file that FOIA by Friday. The network planted 200 trees this month and your city contributed 12. Suggest: organize a planting day." It detects when you are falling off (missed two assemblies, no activity in 14 days) and intervenes with a low-barrier task, not a guilt notification. On a heavy week it suggests less. On a light week it stretches. The agent adapts because it knows your context. It runs on a cheap open-source language model (Llama or Mistral via Ollama) on your own machine. No corporate API required. No data leaves your device unless you choose to share it.

The Revolutionary Calendar is not a Google Calendar with political events scraped from a government website. It is a personalized weekly plan generated from three sources: your own activity log (what you committed to, what is overdue, what is coming up), the network board (what others are doing that you could join or support), and your class progression (what actions advance your skills and standing). The calendar is push, not pull. It arrives via your preferred channel: Matrix message, Signal notification, email digest, or iCal feed. You do not go looking for things to do. The things that need doing come to you, prioritized by urgency and fit.

Each Comrade Agent publishes anonymized activity summaries to a shared Network Board: class, action type, count, and date. No names, no personal data, no location more specific than city. The board is a simple shared data store: a Matrix room, a JSON API endpoint, or a shared file. Every agent reads the board to understand what the network needs. A Scout sees that no FOIA requests have been filed this month and suggests one. An Eco-Warrior sees that the tree planting goal is behind and organizes a day. A Medic sees high activity and no care circles scheduled and proposes one. The Network Board transforms isolated individual action into coordinated collective work without any central coordinator, any surveillance infrastructure, or any authority deciding who does what. The agents decide together, through the emergent logic of gaps and capabilities.

The Comrade Agent is not a product. It is not a service. It is software that runs on your hardware, trained on open-weight models, storing your data locally. You can export your activity log. You can delete it. You can modify the agent's behavior. You can turn it off. The assembly votes on the Network Board's parameters: what activity types are shared, how long data is retained, what the agent can and cannot suggest. The agent serves the participant and the participant serves the assembly. If any corporation ever controls the inference layer, the system has failed.

The Comrade Agent runs on a tiered inference model. Ninety percent of its work is template-based: calendar formatting, streak counting, notification text. No language model needed. Eight percent uses a small local model (Llama 3.1 8B or Mistral 7B via Ollama) for task matching and personalized suggestions. Two percent escalates to a cheap API model (Claude Haiku) for complex planning. Self-hosted cost: $0/month on any modern laptop or a $35 Raspberry Pi 5 running Llama 3.2 3B. API-assisted cost: approximately $0.03 per participant per month. A community of 500 people runs the entire agent infrastructure for less than the cost of a single Zoom subscription.

The Platform working group builds the Civic Engine and writes deployment guides. But individual assemblies need someone to actually set up their Matrix server, deploy their Ghost site, configure their Decidim instance, install Ollama, troubleshoot their Meshtastic nodes, and train their members to use the tools. That is not a working group task. That is an ongoing service. The PRAXIS Technical Cooperative is a worker-owned cooperative that provides this service: the permanent technical infrastructure team for the entire network. It is not a vendor. It is not a contractor. It is a cooperative owned by its workers, governed democratically, accountable to the assemblies it serves, and funded by the assemblies that use it.

Deployment. Set up the full PRAXIS stack for a new assembly: Ghost site, Matrix server, Decidim or Loomio instance, Comrade Agent with Ollama, Meshtastic node configuration, Umami analytics, backups, SSL certificates, DNS. A new city goes from zero to operational infrastructure in one week. Hosting. Managed hosting on cooperative-owned servers for assemblies that cannot or do not want to self-host. No corporate cloud dependencies. Data stays on servers the cooperative controls, under terms the assemblies govern. Support. Help desk for day-to-day issues: password resets, tool configuration, troubleshooting, upgrades. A real person who answers when something breaks, not a ticket queue that disappears into a void. Training. Workshops for assembly Techies: how to administer your own infrastructure, how to deploy Ollama, how to configure the Comrade Agent, how to run backups, how to respond to a security incident. The goal is to make the cooperative's own services unnecessary for each assembly over time. Agent development. Building and maintaining the Comrade Agent: model selection, prompt engineering, Network Board infrastructure, calendar generation, integration with the tool stack. This is the one piece of original software PRAXIS builds, and the cooperative is where the developers work.

The cooperative is owned by its workers: the developers, sysadmins, trainers, and support staff who do the work. One member, one vote. Surplus is distributed as patronage dividends or reinvested by member vote. The cooperative reports quarterly to the PRAXIS network and its pricing is approved by the assemblies it serves. No assembly is required to use the cooperative. Any assembly can self-host, hire someone else, or do it themselves. The cooperative exists because most assemblies will not have a dedicated Techie in their first year, and the movement cannot afford to lose cities because nobody could figure out how to configure a reverse proxy. The cooperative charges on a sliding scale based on assembly size: free for assemblies under 25 members, cost-recovery for 25-100, and sustainable pricing above 100. Grant funding subsidizes the first two years of operation.

The complete tool stack is documented in Section 04 (Architecture) and Section 06 (Technology). The essential tools for a minimum viable assembly, in deployment order: