Home Learn Passport Live MVP Rust Funding Community Blog Papers GitHub ↗
Rust Foundation

Built in Rust for security-critical AI infrastructure

The implementation language should reflect the security philosophy of the project.

Argorix Lang is built in Rust because infrastructure for AI safety should start from a secure systems foundation: memory safety, performance, strong typing and explicit control.

Hardware
Rust Core
Argorix Compiler
Bytecode VM
Execution Traces

Memory Safety

Reduced classes of low-level bugs without relying on a garbage collector.

Performance

Suitable for compiler, VM and bytecode execution paths.

Explicit Control

Clear ownership, predictable behavior and systems-level precision.

Strong Typing

Better foundation for semantic validation and structured runtime logic.

Systems-Level Trust

The implementation stack reflects the project's security-first design.

Language Preview

Designed for AI-agent systems

A language surface for tools, models, permissions, traces and controlled execution.

agent_flow.argx argx compiled · UTF-8
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
tool WebSearch {
  capability: "network.search"
  approval: required
}

model Reasoner {
  provider: simulated
  capability: "analysis"
}

agent Researcher {
  input task

  authorize WebSearch
  authorize Reasoner

  guard verify_sources
  trace enabled

  call WebSearch with task -> results
  ask Reasoner with results -> answer

  return answer
}

Language concepts

  • Tools as explicit capabilities
  • Models as declared providers
  • Authorization blocks
  • Runtime traces
  • Verifiable execution paths
  • Designed to evolve toward sandboxing
runtime output
trace.requested
trace.allowed
tool.call.executed
model.ask.simulated
result.recorded

Syntax and semantics are still evolving. The goal is to make agent behavior easier to reason about, inspect and control.

Quickstart

Build, verify, run

A stable Rust toolchain is all you need. Compile a contract, verify its bytecode, and export evidence — entirely offline.

01

Build & verify the toolchain

Compiles argorixc, argorix-vm and the conformance runner.

02

Compile & run a contract

Validate semantics, emit + verify bytecode, dry-run the VM and export the evidence bundle.

03

Run the governed chat demo

A Next.js chatbot that shells out to the binaries — plan-only by default, never contacting a provider.

argorix — zsh
# build + test the workspace
$ cargo build --workspace
$ cargo test  --workspace

# validate + verify a contract
$ argorixc check examples/runtime_mvp_v100.argx
$ argorixc verify-bytecode examples/runtime_mvp_v100.argx

# dry-run the VM, export evidence
$ argorix-vm run runtime_mvp.argbc.json \
    --dry-run --security-report report.json \
    --evidence-bundle bundle.json

Argorix Lang compiler v1.0
Status: OK
Module: ArgorixChatbotRuntime
Capabilities: 1  Types: 3  Agents: 2  Protocols: 1
Semantic checks: passed

Argorix Bytecode verification v1.0
Status: OK
Bytecode version: 1.0  Instructions: 17
Agents: 2  Protocols: 1

Every command is offline and deterministic. No network access, secrets, environment variables, real tools or models are used to compile, verify or dry-run a contract.