Llm Safety Python Packages

nemoguardrails

NeMo Guardrails is an open-source toolkit for easily adding programmable guardrails to LLM-based conversational systems.

deepteam

DeepTeam is a framework to red team LLMs and AI agents.

agent-airlock

A deny-by-default contract & type-checker layer for AI agent tool calls — Pydantic-based, in-process, zero-core-deps. Validates the actual tool-call payload (ghost-arg stripping, strict types, self-healing retries) beneath MCP gateways & firewalls. Works with LangChain, OpenAI Agents SDK, PydanticAI & CrewAI.

styxx

The measurement layer for machine minds. Reads what a model means and whether it holds the truth; certifies every claim re-runs. meaning_diff + OATH certify + mind profiles + live grounding signal + the cognometric instruments. No torch, no LLM in the loop for the core; MIT, open at the core.

agentguard-governance

Governance layer for autonomous AI agents — pre-flight checks, runtime monitoring, and post-session reporting.

agent-audit

Static security scanner for LLM agents — prompt injection, MCP config auditing, taint analysis. 51 rules mapped to OWASP Agentic Top 10 (2026). Works with LangChain, CrewAI, AutoGen.

uqlm

UQLM: Uncertainty Quantification for Language Models, is a Python package for UQ-based LLM hallucination detection

neurosym-ai

Neuro-symbolic guardrails for LLMs — injection detection, harm filters, output guards, streaming safety, and action-plan validation.

qwed

A deterministic verification layer for AI systems. QWED verifies AI outputs using mathematics, symbolic reasoning, and formal methods (Z3, SMT, SymPy), creating an auditable trust boundary for agentic AI. Not generation. Verification.

yamtam-engine

Audits your AI coding agent setup before it can damage your repo. Scan first. Guard later.

qwed-finance

Deterministic verification middleware for banking and financial AI. NPV, IRR, loan amortization, and interest calculations with QWED precision.

lintlang

lintlang is a static linter for AI agent configs, tool descriptions, and system prompts that runs zero-LLM quality gating in CI. Catches language-level failures (vague tool descriptions, missing stop conditions, schema gaps) before they reach runtime, with deterministic regex + structural detectors and no model calls.

geh

One command to benchmark AI guardrails and coding agents across safety, security, jailbreak, prompt-injection, and secure-code tasks.

kitelogik

Governance middleware for AI agents — policy enforcement, credential lifecycle, audit trail.

fivedrisk

Per-action AI agent risk scoring and governance. Deterministic 5D scoring, HITL gating, FinOps, Agent Cost Management, Markov drift, audit log. Apache-2.0.

gateguard-ai

A fact-forcing hook gate for Claude Code. Makes the AI pause and investigate before editing.

brix-protocol

Runtime Reliability Infrastructure for LLM Pipelines — enforce deterministic rules, measure the Balance Index, and audit every decision.

csl-core

CSL-Core: Deterministic Safety Layer for Probabilistic AI Systems

pytest-wardenbot

Pytest plugin for testing chatbots and LLM apps — prompt injection, jailbreaks, system-prompt leaks, hallucinations, brand drift.

qwed-tax

The Verification Gate for AI-Generated Tax Decisions

blackwall-llm-shield-python

Security middleware for Python LLM apps and services. Blocks prompt injection, masks PII, inspects outputs, and gates agent tools.

qwed-infra

Deterministic verification layer for infrastructure as code. Verifies AWS IAM policies, network reachability, and cost estimates before deployment — sits between AI agents and cloud execution. Powered by Z3 and NetworkX.

agi-pragma

AI Action Firewall — seven-stage Decision Intelligence Core for safe agentic AI

jailbreakeval

[NDSS'25 Best Technical Poster] A collection of automated evaluators for assessing jailbreak attempts.