🚀 Civicnet MCP Server – Model Context Protocol

Welcome to the official open-source MCP Server.

This is the heart of Konstellation & CivicNet’s federated civic infrastructure—a modular, principle-driven server for running local, trusted, and agentic community AI.

---

🧭 What Is the MCP Server?

The MCP Server powers neighborhood, city, or org-level “nodes” in the CivicNet/Konstellation grid. Each node:

• Ingests and validates local data (GIS, policy, community stories, public records)
• Hosts and orchestrates agentic models (reasoning, self-audit, simulation, role-play, etc.)
• Applies civic principles (equity, truth, privacy, transparency) to every query and output
• Publishes secure, documented APIs for dashboards, GIS clients, and external partners
• Federates with other MCP nodes to share knowledge while keeping local control

Why?

So every community can own, govern, and continually improve their own civic intelligence—no black boxes, no vendor lock-in, no silent bias.

---

🌌 Key Features

• LLM-Agnostic: Plug in any modern language model (OpenAI, Anthropic, open source)
• Cloud-Agnostic: Deploy on your cloud, on-prem, edge, or local device
• Principle Repository: Define, audit, and evolve community principles in code
• Modular Agentic Logic: Supports chain-of-thought, self-reflection, role simulation, multi-agent debate, and more
• Data Stewardship: Fine-grained connectors for local, open, and federated civic data
• Simulation-Ready: Run “what-if” scenarios and participatory simulations
• Full Audit Trail: Every output, agent step, and data source is logged and reviewable
• Dockerized: Easy container setup, local or distributed

---

🗃 Directory Overview

---

⚡️ Quickstart (Local Docker Compose)

1. Clone the Repo:

git clone https://github.com/PublikPrinciple/civicnet-mcp-server.git
   cd civicnet-mcp-server

2. Copy and edit your environment variables:

cp config/.env.example .env
   # Edit .env with your API keys, DB, LLM settings, etc.

3. Start the MCP Server (plus optional data services):

docker-compose up --build

4. View API Docs:

• Go to http://localhost:4000/docs (Swagger/OpenAPI by default)

---

🛠️ Core Concepts

Principle Repository

• All outputs are checked, filtered, or rewritten to align with your community’s core principles (e.g., “ensure equity,” “avoid harm,” “cite sources”).
• Update principles via YAML/JSON in /src/principles/.

Agents & Agentic Logic

• Chain-of-Thought Agent: Step-by-step logic for transparency
• Self-Reflective Agent: Checks its own bias, logic, and compliance
• Role Simulation Agent: Simulates debates (planner vs. resident), expert panels, or historical perspectives
• Multi-Agent Collaboration: Supports scenario testing, participatory budgeting, and red teaming
• Each agent’s logic is modular and composable (see /src/agents/).

Data Layer

• Connectors for GIS, census, local CSVs, APIs, and more
• Validation/Redaction steps for privacy and data hygiene
• Federation: Share/selectively with trusted nodes (opt-in)

Prompt & Scenario Templates

• All agentic reasoning is driven by prompt templates (see /src/prompts/)
• Write your own or use built-in civic case study and simulation templates

Simulation

• Model “what if?” futures (housing, climate, budget, etc.) using real and synthetic data
• Outputs are interactive and principle-audited

---

🔑 Example Use Cases

• Generate a “Durham Housing Crisis” case study with maps, timelines, and equity analysis, ready for GIS and public review
• Support a participatory budgeting simulation with role-played debate between community agents
• Publish a plain-language, principle-aligned data API for local dashboards
• Audit all outputs for bias, error, and community principles (with logs and feedback)
• Federate with other neighborhoods/cities to compare, remix, and improve local analyses

---

🧑‍💻 For Developers

• Languages: TypeScript/Node.js by default, Python agent compatibility in roadmap
• Contributing: See CONTRIBUTING.md for branching, code style, and principle-alignment PR checks
• Tests: npm test (Jest), or run all with docker-compose -f docker-compose.test.yml up

---

🏛️ For Civic Data Stewards

• Principle-Driven Governance: Edit /src/principles/ to update your node’s values
• Logs & Audit: Every answer is traceable (who asked, what data, which agents, what checks)
• Open APIs: Build your own GIS client, dashboard, or mobile app on top of the MCP Server’s endpoints

---

🔒 Security & Privacy

• Sensitive data is never shared or exposed without explicit principle-driven consent and redaction
• Each node is sandboxed; federation is always opt-in and auditable

---

🌍 Federation & Constellation

• Konstellation Ready: This server is designed to federate—publish/share insights with other communities, cities, or regions while keeping full local governance and privacy
• Federated queries: Allow comparisons (e.g., “Show eviction rates across all neighborhoods”)

---

📖 Docs & Community

• Full docs in /docs/ or link to civicnet.org/docs
• Join the #mcp-server channel on Discord for support and discussion

---

📝 License

AGPL 2.0 – Open, remixable, and community-driven

See LICENSE for details

---

✨ Get Started Building the Future of Civic Infrastructure!

---

📘 Swagger/OpenAPI Setup (API Endpoint Overview)

Add this to /src/api/routes.ts (sample):

import express from 'express';

import swaggerUi from 'swagger-ui-express';

import YAML from 'yamljs';

const router = express.Router();

const swaggerDocument = YAML.load('./config/swagger.yaml');

router.use('/docs', swaggerUi.serve, swaggerUi.setup(swaggerDocument));

export default router;

Sample swagger.yaml starter:

openapi: 3.0.0

info:

title: MCP Server API

version: 1.0.0

paths:

/api/analyze:

post:

summary: Analyze civic query using agentic logic

requestBody:

required: true

content:

application/json:

schema:

type: object

properties:

query:

type: string

context:

type: object

responses:

'200':

description: Analysis complete