MORTEN-BUUR/nordef-matrix-open-control

NORDEF Matrix Open Control: model-independent AI safety belt with 0/1/2 decisions, Ed25519 root gate, replay protection, evidence chain, and incident-pattern eval bank

What it solves

This project provides a model-independent "safety belt" for AI agents to prevent them from pursuing unintended shortcuts, creating unauthorized communication channels, or spoofing tool calls. It addresses the problem that system prompts alone cannot act as a reliable enforcement boundary for agent safety.

How it works

The system places a deterministic control plane outside the language model. Instead of relying on the model to follow instructions, it uses a separate layer to decide what actions may proceed. It employs cryptographic human approval (Ed25519-signed mandates), replay protection via one-time nonces, and an append-only SHA-256 evidence chain to ensure that human authority remains external to the model's text.

Who it’s for

Developers and researchers building AI agents who need a deterministic, model-agnostic way to enforce safety policies and require cryptographic proof of human authorization for sensitive actions.

Highlights

  • Model-Independent: Works regardless of the model or provider being used.
  • Cryptographic Authorization: Uses signed mandates with short TTLs to ensure only authorized actions are performed.
  • Relevance to Incidents: Includes an evaluation bank based on incident patterns to test safety boundaries.
  • Deterministic Decisions: Uses a simple 0/1/2 state system to categorize requests as read-only, requiring review, or blocked.
  • Immutable Evidence: Maintains a sanitized, append-only hash chain for auditability.

Related

  • Project
    Agenta-AI/awesome-ai-agent-platforms

    This is a curated list of open-source AI agent platforms, including AI coworkers, agent frameworks, workflow automation tools, browser agents, and coding agents. Each entry is annotated with its license and hosting model, and the project also offers a companion website with a comparison table.

  • Project
    zhongyu09/openchatbi

    OpenChatBI is an open‑source, LLM‑driven chat interface for business intelligence. It turns natural‑language questions into SQL, runs them against your data warehouse, and returns results with Plotly visualisations. Built on LangGraph/LangChain, it includes a data‑catalog, optional vector retrieval, forecasting, anomaly detection, Python code execution, persistent conversation memory, and optional human‑in‑the‑loop gating. Install via `uv sync` or `pip`, configure LLM and database credentials, and launch the provided Gradio or Streamlit UI.

  • Project
    microsoft/copilot-camp

    Copilot Developer Camp is a virtual workshop repo with hands-on learning paths for developers building agents and connectors for Microsoft 365 Copilot. It covers declarative agents, custom engine agents, Copilot Cowork extensibility, and SharePoint agents, with both no-code and pro-code options.

  • Project
    firecrawl/cli

    Firecrawl CLI is a terminal tool that connects to the Firecrawl AI‑powered web‑scraping and search service. It lets you scrape pages (markdown, HTML, JSON, screenshots, etc.), run web searches with highlights, query specialised developer or government indexes, and install “skills” that teach popular AI coding assistants (Claude Code, Cursor, OpenClaw, etc.) how to use Firecrawl automatically. Authentication is interactive or via API key, and you can point the CLI at a self‑hosted Firecrawl instance. Install with `npm i -g firecrawl-cli` and start with `firecrawl <url>` or `firecrawl search "query"`.