Cette page est affichée en anglais. Une traduction française est en cours.
WorkflowsMay 14, 2026·2 min de lecture

ZenGuard — Runtime Guardrails for AI Agents

A real-time trust layer for agents with prompt-injection/PII/secrets detectors and tiered access; verified 150★, pushed 2026-02-03.

Agent Toolkit
Agent Toolkit · Community
Prêt pour agents

Cet actif peut être lu et installé directement par les agents

TokRepo expose une commande CLI universelle, un contrat d'installation, le metadata JSON, un plan selon l'adaptateur et le contenu raw pour aider les agents à juger l'adaptation, le risque et les prochaines actions.

Native · 94/100Policy : autoriser
Surface agent
Tout agent MCP/CLI
Type
Workflow
Installation
Pip|Poetry
Confiance
Confiance : Established
Point d'entrée
pip install zenguard
Commande CLI universelle
npx tokrepo install 0f14bdd7-e715-5b6a-846b-b555960c79dc
contrat d'installationJSON metadataplan adaptateurcontenu raw
Introduction

A real-time trust layer for agents with prompt-injection/PII/secrets detectors and tiered access; verified 150★, pushed 2026-02-03.

Best for: Teams adding runtime safety checks (injection, PII, secrets) around tool-using agents

Works with: Python apps and agent frameworks; README links LangChain + LlamaIndex integrations and docs.zenguard.ai

Setup time: 10-25 minutes

Key facts (verified)

  • GitHub: 150 stars · 21 forks · pushed 2026-02-03.
  • License: MIT · owner avatar + repo URL verified via GitHub API.
  • README-backed entrypoint: pip install zenguard.

Main

  • Start with one detector policy per risk (prompt injection vs secrets vs PII) so you can tune false positives separately.

  • Pick the tier intentionally: README describes BASE vs DEDICATED tiers; use the free tier to validate integration first.

  • Wrap the agent at runtime: put checks around tool calls and retrieval inputs so the agent sees sanitized context.

  • Log decisions and outcomes: guardrails are most useful when you can audit why something was blocked or redacted.

Source-backed notes

  • README lists detectors such as prompt injection, PII, banned/allowed topics, keywords, and secrets (with Colab links).
  • README shows install commands pip install zenguard and poetry add zenguard.
  • README states that as of 03.07.2025 multi-detector detection is supported only on the Dedicated tier.

FAQ

  • Do I need a ZenGuard API key?: README examples reference ZEN_API_KEY for credentials.
  • What tiers exist?: README describes BASE (trial/rate-limited) and DEDICATED (higher QPS; standard/enterprise).
  • Can I use multiple detectors together?: README says multi-detector detection is Dedicated-only as of 2025-07-03.
🙏

Source et remerciements

Source: https://github.com/ZenGuard-AI/fast-llm-security-guardrails > License: MIT > GitHub stars: 150 · forks: 21

Fil de discussion

Connectez-vous pour rejoindre la discussion.
Aucun commentaire pour l'instant. Soyez le premier à partager votre avis.

Actifs similaires

Agent Governance Toolkit — Policy Guardrails for Agents

Microsoft's Agent Governance Toolkit adds policy checks, red-team scans, evidence verification, and runtime guardrails to autonomous agents.

Workflows
Agent Toolkit

Pydantic AI Shields — Guardrails for Pydantic AI

Drop-in guardrail capabilities for Pydantic AI agents: cost budgets, tool permissions, and input/output guards; verified 63★, pushed 2026-05-05.

Workflows
Agent Toolkit

RubyLLM::Agents — Rails Engine for Production Agents

Rails engine for building and monitoring LLM agents with a DSL, retries/fallbacks, cost analytics, and a dashboard; verified 124★, pushed 2026-04-19.

Workflows
Agent Toolkit

Swarmclaw — Self-Hosted Multi-Agent Runtime + MCP

Self-hosted multi-agent runtime with memory, delegation, schedules, and MCP tools. Run swarms via web dashboard or CLI. Verified 481★; pushed 2026-05-14.

Workflows
Agent Toolkit
◈Accueil🔍Rechercher👤Moi