This repository contains the public documentation for QWED — a deterministic verification ecosystem that proves, validates, and enforces trust boundaries for AI systems, agents, infrastructure, and computational workflows.
The documentation covers:
| Area | Description |
|---|---|
| QWED Verification Protocol | Core verification engines (Math, Logic, SQL, Code, Facts) and the Untrusted Translator architecture |
| AI & Agent Security | Agentic security guards — MCP Poison, RAG, Exfiltration, Process Determinism |
| Finance Guards | NPV/IRR verification, loan amortization, compound interest, ISO 20022, currency safety |
| Legal Guards | Deadline verification, clause consistency, citation validation, jurisdiction checks, statute of limitations |
| SDK References | Python, TypeScript, Go, and Rust SDK documentation and API references |
| Architecture | Deterministic verification design, fail-closed philosophy, verification boundaries |
| Deployment | Docker, Kubernetes, CI/CD integration, and operational guidance |
| Package | Description | PyPI | npm | Repo |
|---|---|---|---|---|
| qwed | Core verification + security guards | qwed |
@qwed-ai/sdk |
qwed-verification |
| qwed-finance 🏦 | Banking, loans, NPV, ISO 20022 | qwed-finance |
@qwed-ai/finance |
qwed-finance |
| qwed-legal ⚖️ | Contracts, deadlines, citations, jurisdiction | qwed-legal |
@qwed-ai/legal |
qwed-legal |
| qwed-tax 💸 | Tax compliance & withholding verification | qwed-tax |
— | qwed-tax |
| qwed-infra ☁️ | IaC verification (Terraform, IAM, Cost) | qwed-infra |
— | qwed-infra |
| qwed-ucp 🛒 | E-commerce cart/transaction verification | qwed-ucp |
— | qwed-ucp |
| qwed-mcp 🔌 | Claude Desktop MCP integration | qwed-mcp |
— | qwed-mcp |
| open-responses 🤖 | OpenAI Responses API + QWED guards | qwed-open-responses |
— | qwed-open-responses |
- Node.js (v18+)
- Mintlify CLI
npm i -g mintlifymintlify devPreview locally at: http://localhost:3000
/
├── legal/ # QWED-Legal guard documentation
├── finance/ # QWED-Finance guard documentation
├── tax/ # QWED-Tax guard documentation
├── sdks/ # SDK reference (Python, TypeScript, Go, Rust)
├── open-responses/ # OpenAI Responses API integration docs
├── specs/ # Verification specification documents
├── snippets/ # Shared reusable MDX snippets
├── images/ # Static assets
├── changelog.mdx # Release changelog
├── architecture.mdx # System architecture
├── faq.mdx # Frequently asked questions
├── mint.json # Mintlify configuration
├── LICENSE # Apache License 2.0
└── README.md
QWED is built around deterministic verification principles:
- Fail closed by default — unknown states are never treated as safe
- Verification requires explicit evidence — no silent assumptions, no fabricated defaults
- Heuristics must not silently become proof — LLM outputs are always marked
HEURISTICwhen symbolic verification is not possible - Auditability and reproducibility are first-class — every verification is deterministic and inspectable
- The LLM is the Untrusted Translator — it converts natural language to formal DSLs; symbolic engines are the Trusted Verifiers
The documentation follows the same philosophy:
- Precise over promotional
- Explicit over ambiguous
- Operational over hype
Probabilistic systems should not be trusted with deterministic tasks. If it can't be verified, it doesn't ship.
Contributions are welcome for:
- 📝 Documentation fixes and typo corrections
- 🔧 Broken code examples
- 💡 Clarification requests
- 🐛 Inaccurate guard behavior descriptions
Please open an issue or submit a pull request. For contribution guidelines, see the main QWED Contributing Guide.
If you discover a security issue related to QWED documentation, examples, or verification workflows, please follow the repository security policy.
Copyright 2025 QWED-AI
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0