developmentAI-assistedtransparency
One Developer, 250,000 Lines, and a Lot of Help
Johnathon Rhoades
AI agents wrote most of the code. Every line was governed. Here's what that means and why it matters.
TVN — Truth by Verified Node
AI Tools That Behave.
By Design.
TVN Bridge is a local-first governance platform that enforces constraints AI assistants cannot bypass. Your code. Your rules. Your machine.
A neurosymbolic governance architecture — AI handles reasoning, deterministic rules enforce boundaries the AI cannot prompt-engineer past.
The TVN Ecosystem
Independent products that combine. Each works alone. Together, they create an air gap between AI and your data.
Product
What You Get
What AI Can Access
Encrypted local workspace — notes, databases, accounting. SQLCipher at rest, zero telemetry. Works standalone.
Nothing. Data is encrypted.
TVN Bridge
AI governance framework — constraints, truth nodes, context gates, audit trails. Governs any AI agent. Works standalone on any codebase.
Only what truth nodes and constraints allow.
Vault + Bridge
Governed AI access to your encrypted data over TSP. Lock the vault, AI goes cryptographically blind. Four layers of "no" before AI sees anything.
Governed, redacted, audited access only.
TSP: Mutual HMAC auth + AES-256-GCM encrypted IPC. Lock the vault, AI goes blind. Cryptographically. Read the protocol breakdown →
Pluggable enforcement: TVN's architecture accepts third-party policy engines (Microsoft AGT, AEGIS) as components beneath its knowledge and orchestration layer. We're the operating system. They're the drivers.
TVN in Action
[ Demo video — record Sovereign Console ]
How It Works
Install
pip install tvn-bridge
Index Your Project
tvn index — semantic map in seconds
Define Rules
Truth Nodes + Constraints
AI Works Under Governance
Firewall enforces every action
Review
Audit trail — JSON/CSV export
What TVN Bridge Does
🛡
Sovereign Firewall
3-tier security blocks 30+ destructive operations in <10ms. OWASP LLM06 compliant.
Proprietary🔎
Context Gate
90-97% token savings, enforced. Forces AI to search before reading. No known prior art.
Proprietary📜
Truth Nodes
Human-defined invariants injected into every AI context. AI cannot override them.
👁
Universal Audit Trail
Every ALLOW/BLOCK decision logged. Export as JSON/CSV. Know exactly what AI did.
🏠
Local-First
Zero network calls. Zero telemetry. Your data never leaves your machine.
🔌
MCP Integration
17 governance tools for Claude, Cursor, Copilot, Gemini, and more. One-click install. Works with any MCP-capable agent.
Works with every major AI agent
Claude Code Claude Desktop Cursor VS Code / Copilot Gemini Grok Any CLI via tvn wrap
Tier 1 (full governance) · Tier 2 ( MCP governance) · Tier 3 (env + audit)
Vendor lock-in prevention: your governance rules, agent memory, and collaboration patterns live in TVN, not in any AI vendor's cloud. Switch models without losing institutional knowledge.
Before & After
Without TVN
With TVN
AI reads 40 files to find what it needs
AI searches semantically, reads 3 files
31,000+ tokens consumed per question
1,600 tokens consumed per question
Redundant reads every session
Persistent memory across sessions
AI "forgets" and re-explores constantly
Truth Nodes + Memory Nodes eliminate repetition
No audit trail
Every action logged, exportable as JSON/CSV
No file write governance
Every write requires human confirmation
95% less waste per interaction
Context Gate eliminates redundant file reads and token consumption. For a team of 5, that translates to ~$2,700/year in reduced API costs.
Estimate based on typical Claude API usage at current pricing.
Who It's For
Solo Developers
Stop burning tokens on files AI doesn't need.
You pay per token. Context Gate cuts that by 90-97%.
Enterprise Teams
Audit trail for every AI action. Export as JSON/CSV.
Compliance teams need proof. TVN provides it automatically.
Security-Conscious Orgs
OWASP LLM06 compliant. Red-team validated.
Firewall blocks 30+ destructive operations in <10ms.
Multi-Agent Teams
One governance layer for Claude, Cursor, Copilot, and more.
Tool name aliases + governance adapters work cross-agent.
CTOs & Tech Leads
Human authority over AI, enforced by architecture.
Truth Nodes + Zero Trust = controllable amplification, not blind trust.
90-97%
Token Savings
<10ms
Firewall Latency
1,494
Tests Passing
Zero
Telemetry
See It In Action
Dashboard Overview
Firewall Activity
Smart Core / Search
Agent Sessions
Launch Pad
Frequently Asked Questions
What is TVN Bridge? ▼
TVN stands for Truth by Verified Node. TVN Bridge is a local-first AI governance platform. It sits between you and your AI coding assistant, enforcing rules the AI cannot bypass — like a firewall for AI behavior. It controls what files AI can access, what operations it can perform, and how much context it consumes.
Does it send my code to the cloud? ▼
No. TVN Bridge is 100% local. Zero network calls, zero telemetry, zero data collection. Your code never leaves your machine. The AI assistant you use (like Claude) has its own connection, but TVN Bridge itself phones home to nobody.
How does it save 90-97% on tokens? ▼
The Context Gate blocks AI from reading files directly until it searches first. Instead of grepping your entire codebase and dumping thousands of lines into context, the AI is forced to use semantic search to find only the relevant code. The result is dramatically smaller context windows and faster, more accurate responses.
What AI tools does it work with? ▼
TVN Bridge integrates with Claude Code, Claude Desktop, Cursor, VS Code / Copilot, Gemini, and Grok via the Model Context Protocol (MCP). It exposes 17 governance tools that enforce rules regardless of which AI agent is calling them. Any CLI tool can also be wrapped via tvn wrap for environment injection and audit trails.
Is it free? ▼
TVN Bridge is stewarded by The Rhoades Institute of Technology, a 501(c)(3) nonprofit. We’re committed to making AI governance accessible. Expect a free tier for individuals and fair pricing for teams. Join the waitlist to be the first to know.
Who is behind this? ▼
TVN Bridge is built by Johnathon Rhoades and stewarded by The Rhoades Institute of Technology, a 501(c)(3) nonprofit dedicated to transparent, accountable AI tooling. 250,500 lines of code across 9 projects and 7 languages, built by one developer — nights and weekends while working full-time, until January 2026.
"But you still use cloud AI (Claude, GPT)" ▼
Yes — TVN acts as a governance filter between you and the API. It controls what data reaches the cloud, blocks sensitive files, and limits token consumption. It’s not anti-cloud. It’s pro-control.
"95% token savings sounds too good to be true" ▼
Measured, not estimated. The Context Gate forces AI to search semantically before reading files. Instead of reading 40 files, it reads 3. The math is straightforward.
"Doesn’t governance slow AI down?" ▼
The Firewall adds <10ms per decision — invisible. Meanwhile, the Context Gate saves 83% of tool calls, so governed AI is actually faster than ungoverned AI.
"Won’t AI companies just build this themselves?" ▼
They profit from more API calls, not fewer. A 95% token reduction is a 95% revenue reduction for them. They won’t build this. TVN’s enforcement layer is designed with a pluggable backend architecture — enforcement engines like Microsoft’s Agent Governance Toolkit and AEGIS can be integrated as components beneath TVN’s knowledge and orchestration layer. We’re the operating system. They’re the drivers.
"This is just for developers, right?" ▼
Today, primarily. The architecture is domain-agnostic. We already support code, personal assistant, and tabletop gaming domains. TVN Vault extends the governance model into encrypted database management for anyone. The governance model applies to any AI use case.
By the Numbers
Lines of code 250,500
Projects 9
Languages 7
Tests passing 1,494
Governed agent sessions 121+
Developers 1
How It Started
TVN started because ChatGPT couldn't follow the rules of a tabletop RPG. If AI can't remember game rules, it can't be trusted with your codebase.
Built nights and weekends while working full-time — until January 2026. Designed and directed by one developer. Code written by AI agents from Anthropic, OpenAI, and Google, orchestrated under human authority.
The architecture implements neurosymbolic governance — AI agents handle reasoning (neural), while truth nodes, constraints, and context gates enforce rules deterministically (symbolic). The AI can't prompt-engineer its way past a symbolic rule.
Built for the Public Benefit
TVN is developed by The Rhoades Institute of Technology (RIT), a Wisconsin 501(c)(3) nonprofit dedicated to ethical technology education. This isn't a VC play — it's a civic mission.
We believe AI should serve humans — not revenue models, not engagement metrics, not ad networks. Every tool we build is designed to be inspectable, transparent, and accountable.
Built with open source: Ralph autonomous loop pattern (Geoffrey Huntley + community), Loopy causal diagrams (Nicky Case, CC0). Designed to integrate MIT/Apache-licensed enforcement engines.
Ready to take control?
TVN Bridge is in active development. We’re presenting at CypherCon 2026 (April 1–2, Milwaukee) and Midwest Gaming Classic (April 24–26). Join the waitlist for early access, or find us at the event.
v0.11.0 Alpha · Current as of March 2026