Skip to content
| Marketplace
Sign in
Visual Studio Code>Linters>Trinetra — Architectural GuardrailNew to Visual Studio Code? Get it now.
Trinetra — Architectural Guardrail

Trinetra — Architectural Guardrail

utsarga dhakal

|
1 install
| (0) | Free
Local architectural guardrails with explainable evidence for TypeScript, JavaScript, and Python. IDE sidebar, headless policy engine, and pre-commit hook — no external APIs.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

Trinetra — Architectural Guardrail

Local architecture governance for TypeScript, JavaScript, and Python — no cloud, no LLM, no external APIs.

Trinetra deep-resolves import graphs (including barrel re-exports up to 8 hops), evaluates changes against declared architecture rules, and produces explainable proof packets with evidence chains. Everything runs on your machine or in CI.

Features

  • Declared architecture — layers, forbidden edges, mandatory gateway paths via .trinetra/architecture.json
  • Deep barrel resolution — catches violations hidden behind export * from chains
  • Policy engine — scoring, block decisions, proof packets, and repair suggestions
  • VS Code sidebar — Workspace Inspector with score, signals, and evidence click-through
  • Headless CLI — trinetra evaluate, pre-commit, merge-check, plan-merge, MCP server
  • Signed receipts — optional Ed25519 architecture-proof/v1 receipts for CI audit trails
  • 100% local — deterministic, zero egress

What's new in 1.3.0

  • Python architecture checks — full import graph analysis for .py files, including __init__.py barrels
  • Polyglot repos — catches violations across Python ↔ TypeScript/JavaScript chains in one scan
  • Workspace-wide scan — analyzes all TS/JS/Python files in your project (footer shows file + violation counts)
  • File extension fallback — works even when the editor mis-detects language (uses .py/.ts/.js extension)

Quick start

1. Install the extension

Install from the VS Code Marketplace, or from a .vsix:

code --install-extension trinetra-1.2.0.vsix

2. Add architecture config

Create .trinetra/architecture.json in your repo root:

{
  "version": 1,
  "mode": "declared",
  "defaultLayer": "generic",
  "layers": [
    { "name": "controller", "match": ["**/controllers/**", "**/pages/**"] },
    { "name": "service", "match": ["**/services/**"] },
    { "name": "repository", "match": ["**/repositories/**"] },
    { "name": "database", "match": ["**/orm/**", "**/database/**"] }
  ],
  "forbidden": [
    { "from": "controller", "to": ["database", "repository"] }
  ]
}

3. Open the Trinetra sidebar

Click the Trinetra icon in the activity bar → Workspace Inspector shows verdicts, signals, and evidence for the current file.

4. Run from CLI (optional)

From the extension folder (or after npm install -g when published to npm):

npm run compile
node bin/trinetra.js evaluate --file src/controllers/Checkout.ts
node bin/trinetra.js pre-commit

MCP for AI agents

Expose Trinetra to Cursor or other MCP clients:

npm run mcp

Agents get deterministic pre-check / post-verify tools (~50ms per file) with full evidence chains for repair loops.

Pre-commit hook

node bin/trinetra.js pre-commit

Exit code 1 blocks the commit when architecture violations are found.

Performance

Stress-tested on synthetic monorepos:

Scale Cold batch Throughput
20k files ~9s ~2,100 files/sec
100k files ~30s ~3,400 files/sec

Run locally: npm run stress or npm run stress:100k

License

Apache-2.0 — see LICENSE and OPEN_CORE.md for the open-core split.

  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
© 2026 Microsoft