Network-AI
Documentation hub

Everything needed to evaluate and run Network-AI.

Installation, first run, CLI and PowerShell usage, architecture and race-condition control, benchmarks, security, enterprise evaluation, audit logging, integration references, AI tool-use files, community policies, and licensing are all collected here.

Start here

Start with the core docs

Start with setup, the control model, and release history.

Step 1

Get to first run quickly

Use the quick start for install, CLI examples, PowerShell flows, and the shortest route to a working orchestrator.

Step 2

Inspect the control model

Review architecture, AuthGuardian, trust levels, and audit schema to understand how state, permissions, and workflow boundaries are enforced.

Step 3

Verify the trust signals

Check the security policy, benchmarks, and changelog together to verify support windows, release cadence, and operational discipline.

Core documentation

Operational and product documents

These are the main files a technical evaluator, implementer, or security reviewer will actually read first.

Getting started

Quick Start

QUICKSTART.md

Installation, first run, multi-framework setup, CLI usage, PowerShell and Python script workflows, and the shortest path from package install to a running orchestrator.

  • Clone, install, and run a working setup check
  • Hello-world orchestration flow and adapter selection
  • CLI commands for blackboard, auth, budget, and audit
  • Project context and companion Python utilities
System design

Architecture

ARCHITECTURE.md

The race-condition problem Network-AI solves, the control-plane components, FSM journey design, delegation handoff protocol, and the structure of the orchestration layers.

  • Atomic propose-validate-commit blackboard model
  • AuthGuardian, FederatedBudget, AdapterRegistry, JourneyFSM
  • FSM stage transitions and compliance violations
  • Handoff protocol and project structure reasoning
Performance

Benchmarks

BENCHMARKS.md

Provider performance, rate-limit behavior, local GPU and self-hosted deployment guidance, and the `max_completion_tokens` sizing rules that prevent silent truncation.

  • Validation throughput and latency ranges
  • Cloud provider response times and RPM tradeoffs
  • Local Ollama or vLLM guidance for zero rate limits
  • Token-cap recommendations by task shape
Security

Security Policy

SECURITY.md

Supported versions, vulnerability reporting, the security module, permission system, trust levels, audit trail, scanning posture, and disclosure policy.

  • AES-256-GCM, signed tokens, rate limiting, and path protections
  • AuthGuardian scoring and default resource restrictions
  • Audit log behavior and token revocation flows
  • CodeQL, Snyk, VirusTotal, and disclosure process
Enterprise

Enterprise Evaluation Guide

ENTERPRISE.md

A fast evaluation path for architects and technical buyers covering offline operation, auditability, security review posture, stability policy, and integration entry points.

  • Quick evaluation checklist and architecture summary
  • Stability signals and support expectations
  • Security and supply-chain posture overview
  • Enterprise integration entry points and examples
Integration

Integration Guide

INTEGRATION_GUIDE.md

An end-to-end integration walkthrough for technical leads, from discovery and agent inventory through phased rollout, framework mapping, validation, and common mistakes.

  • Framework mapping across the supported adapter set
  • Primitive mapping for race control, auth, budgets, and audit
  • Phased rollout path to avoid integration disruption
  • Enterprise concerns and validation checklist
Governance references

Audit, permissions, trust, adapters, and adoption

These files explain how the control plane behaves, how access is scored, how events are logged, and how the ecosystem plugs into the orchestrator.

Audit

Audit Log Schema

AUDIT_LOG_SCHEMA.md

The append-only audit trail specification, including the event envelope, event types, permission request fields, denied-request scoring payloads, and lifecycle events.

  • JSONL file structure and CLI access patterns
  • Permission request, grant, deny, revoke, and cleanup events
  • Budget, handoff, safety shutdown, and context events
  • Weighted score fields recorded for denials
Security reference

AuthGuardian

references/auth-guardian.md

The full permission-wall reference, including weighted scoring, thresholds, token lifecycle, default restrictions, and audit logging behavior for protected resources.

  • Approval score = justification + trust + risk weighting
  • Token structure, expiry, validation, and revocation
  • Restrictions for DATABASE, PAYMENTS, EMAIL, FILE_EXPORT
  • Operational examples for requesting and using grants
Trust model

Trust Levels

references/trust-levels.md

The agent trust model used by AuthGuardian, from default trust levels through trust bands, approval impact, and guidance for introducing new agents safely.

  • Default trust scores for orchestrator, assessor, and analyst agents
  • Trust bands from full trust to automatic denial
  • Worked scoring examples and denial thresholds
  • Options for pre-registration and gradual trust building
Extensibility

Adapter System

references/adapter-system.md

The adapter architecture for plugging in existing frameworks, custom handlers, and mixed agent stacks without locking the system to a single ecosystem.

  • Adapter registry overview and routing design
  • Custom, LangChain, AutoGen, CrewAI, and MCP examples
  • How to wrap existing logic in a governed adapter
  • Guidance for writing custom adapters cleanly
Ecosystem

Adopters

ADOPTERS.md

The public adopters list for organizations and open-source users, including the process for adding yourself through a pull request.

  • Organizations and individual project adopter tables
  • Open invitation to add your usage publicly
  • PR steps and expected row fields
  • Anonymous use-case-only entries accepted
AI integration files

Use with Claude, ChatGPT, and Codex

Three repository-root integration files are already included for tool use, custom GPT actions, and Claude project configuration.

Tool use

claude-tools.json

Drop directly into a Claude API tools array or compatible Codex-style tool-use integration to expose Network-AI operations.

GPT actions

openapi.yaml

Import directly into the GPT editor as a Custom GPT Action spec for swarm delegation, state inspection, and orchestration endpoints.

Claude projects

claude-project-prompt.md

Paste into Claude Projects custom instructions to configure the orchestrator agent behavior without requiring the MCP server.

Project standards

Community, contribution, license, and policy

These files explain how the project is governed publicly and what contribution quality bar the repository expects.

Conduct

Code of Conduct

Professional standards, unacceptable behavior, enforcement scope, and reporting paths for the project community.

Contributing

Contributing Guide

Issue-first contribution workflow, test and type-check requirements, security expectations, documentation updates, and review criteria.

License

MIT License

Permissive licensing with rights to use, modify, and distribute the software, provided the copyright and permission notice stay with the code.

Release hygiene

Security and public trust surface

Use the security policy, advisories, changelog, and benchmark docs together when evaluating maintenance discipline and production readiness.

Additional repo docs

Overview, release notes, examples, and operator guides

These sections cover the rest of the public repository documentation.

Overview

Repository Overview and Examples

README.md · examples/README.md

Read the repository overview and the runnable examples guide in site format.

Release history

Changelog

CHANGELOG.md

Review version history, release cadence, and recorded changes.

AI operator docs

Claude, Codex, and Skill Files

CLAUDE.md · CODEX.md · SKILL.md · skill.json

Operator guidance and machine-readable manifests for Claude, Codex, and skill-based integrations.

Deep references

Blackboard Schema and MCP Roadmap

references/blackboard-schema.md · references/mcp-roadmap.md

Reference material for architecture, state layout, and MCP planning.

Ecosystem

Awesome Lists and Show HN

AWESOME_LISTS.md · SHOW_HN.md

Ecosystem and launch-positioning documents for the broader project story.

These pages present the public repository docs in site format, while the source markdown, JSON, YAML, and license files remain available on GitHub.