🌍 Semantic Runtime · Open Protocol · 2026-Ready

Trasgo turns LLM context into a self-correcting virtual machine.

A self-initializing semantic runtime and context codec (§1) for autonomous LLM reasoning and formal verification. Execute, verify, and evolve logic natively in the context window.

Explore the repo Read the README Open an issue

Live Sandbox

Try the trasgo CLI right in your browser. The web runtime uses Hugging Face Router with a selectable model backend. Run trasgo login to authorize with Hugging Face, then use trasgo models and trasgo model <id> before the live demo.

trasgo — semantic-runtime-shell
Trasgo Web Sandbox · Hugging Face Inference Providers
Type trasgo --help to see commands, or try the demo loop:
1. trasgo init "demo"
2. trasgo login
3. trasgo models
4. trasgo model Qwen/Qwen3-4B-Instruct-2507
5. trasgo boot
6. trasgo eval
7. trasgo send "Reduce (λx.λy.x) y"
 
$

Installation instructions and advanced CLI usage are available in the README.md.

2026 dev stack alignment

Semantic Runtime (LLM-as-VM) Agentic Harness (ERR/CT) Formal Verification (V1-V6) Protocol Evolution (σ-axis) Instruction Set Architecture (§1) Structured JSON mode Model-selectable sandbox

Recorded demos

Trasgo codec demo Trasgo runtime shell demo