Skip to content
@vectimus

Vectimus

Deterministic governance for AI agents. Open source. Sub-10ms. Zero config.

AI coding agents and agentic frameworks can run shell commands, write files, call APIs and invoke MCP tools. Vectimus intercepts every action and evaluates it against Cedar policies before execution. Same input, same decision, every time.

pipx install vectimus
vectimus init

Two commands. Your agents governed before they execute.


Repositories

vectimus The core policy engine. Persistent daemon evaluates Cedar policies in under 10ms with zero telemetry. Works with Claude Code, Cursor, GitHub Copilot, Gemini CLI, LangGraph, Google ADK and the Claude Agent SDK. Apache 2.0.
policies Cedar rules across 11 policy packs: destructive ops, secrets, supply chain, infrastructure, code execution, data exfiltration, file integrity, database, git safety, MCP safety and agent governance. Every rule traces to a real incident and maps to OWASP, NIST, SOC 2 and ISO 27001 controls.
sentinel Autonomous threat-to-policy pipeline. Three AI agents scan the agentic AI threat landscape daily, draft Cedar policies, prove them against reconstructed attacks in a sandbox, then open PRs for human review. Sentinel's own tool calls are governed by Vectimus.

How it works

┌─────────────┐     ┌───────────────┐     ┌──────────────┐     ┌──────────┐
│  AI Agent   │────▶│               │────▶│ Cedar Policy │────▶│ allow /  │
│ (tool call) │     │   Vectimus    │     │   Engine     │     │ deny /   │
│             │◀────│               │◀────│              │◀────│ escalate │
└─────────────┘     └───────────────┘     └──────────────┘     └──────────┘
                           │
                     ┌─────┴─────┐
                     ▼           ▼
              ┌──────────┐ ┌──────────────┐
              │Audit Log │ │Signed Receipt│
              │ (JSONL)  │ │ (Ed25519)    │
              └──────────┘ └──────────────┘
  • Deterministic -- No LLM in the governance loop. Cedar policies, not probability.
  • Local -- Everything evaluates on your machine. No telemetry, no account, no cloud calls.
  • Fast -- Persistent daemon keeps the Cedar engine warm. Sub-10ms per evaluation.
  • Observable -- Cryptographically signed audit receipts for every decision. Start in observe mode, enforce when ready.
  • Incident-driven -- Sentinel discovers real-world attacks and ships verified policies daily.

Compliance coverage

OWASP Agentic Top 10 (all 10 categories) | SOC 2 | NIST AI RMF | NIST CSF 2.0 | ISO 27001 | EU AI Act | SLSA | CIS Controls


Website  •  Docs  •  Blog

Pinned Loading

  1. vectimus vectimus Public

    Deterministic governance for AI coding agents. Cedar-based policy engine that intercepts every agent action and evaluates it against deterministic rules before execution.

    Python 29 7

  2. sentinel sentinel Public

    Automated threat intelligence pipeline for agentic AI security. Three AI agents discover threats, draft Cedar policies and prove they work in a sandbox. Governed by Vectimus.

    Python 1 1

  3. policies policies Public

    78 Cedar policies and 369 rules governing AI coding agents. Every rule traces to a real incident, published CVE, or compliance framework requirement (SOC 2, NIST, ISO 27001, EU AI Act, OWASP).

    JavaScript 1

Repositories

Showing 4 of 4 repositories

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Loading…

Most used topics

Loading…