AI guardrails & agent security
3,625 agents · most starred first
AI guardrails & agent security, page 31 of 37
AI governance SDK — intercept, audit, and control AI agent tool calls. Add human-in-the-loop approval, kill switches, and a full audit trail to any AI agent.
The official command-line interface for Creduent: the open standard and security protocol for cryptographic AI agent identity, authentication, and trust verification using Ed25519 and DNS.
Phone approvals for LangGraph and LangChain agents. Your agent asks, your user taps Approve or Deny.
Framework adapters that wire TryAgent human-in-the-loop guardrails into LangGraph, the Vercel AI SDK, and the OpenAI Agents SDK.
CLI for toolgovern -- validate policy files, audit local gate traces, and scaffold framework integration boilerplate, with a --json output mode for programmatic/agent use.
SusFactor jailbreak/prompt-injection guardrail for LiteLLM
A security firewall for AI agents, skills and MCP servers — scan before you install.
AgentGuard installation entry point
Guard API / Control Plane service for AgentGuard
Headless command-line control utility for AgentGuard
Stateless AgentGuard Core security decision engine
Official Python SDK for Aeonic — runtime monitoring and governance for AI agents
A secure-by-default authorization PEP for APIs, Agent tools, RAG, and tasks
Guardrails for AI-agent tool/function calls: schema validation, realistic error recovery, cost/quota budgets, circuit breakers, and response caching. Framework-agnostic.
Deterministic safety gate for AI coding agents
DefenseClaw evaluators for agent-control
Agent Control Specification: a stateless policy decision runtime implementing the agent-hooks interceptor contract.
KeePass-backed credential vault for agent-core agents
Inbound notifications router — deny-by-default external-event surface for agent-core beings.
Flags risky changes to AI agent configs (prompts, tool permissions, model choice) in a diff.
User-space runtime policy guard with deterministic control decisions and forensic evidence.
Public Preview — Unified installer and runtime policy enforcement for the Agent Governance Toolkit
Embedded multi-tenant Agent governance engine - cost attribution, declarative policies, and audit
Framework-neutral agent control contract: interception points, agent context, verdict, and conformance test kit. Python wrapper over the canonical Rust core. Cooperative contract, not a security boundary: the host is fully trusted (see SECURITY.md).
Python client for agent-input-sanitizer: a thin bridge to the Node.js CLI for byte-identical sanitization verdicts.
Local runtime guard for coding agents with loop detection and explainable policy decisions.
Python SDK for Agent Passport - AI Agent compliance proof system for EU AI Act Art.50
Action Commit Safety for AI agents: exact-call approval, UNKNOWN outcome recording, idempotent execution, and signed audit.
OpenSandbox backend SDK for AI agents
Python SDK for agent-toolbox.ai — deterministic pre-action gates for AI agents
Safety checks an AI agent runs before it acts: is this package malware/typosquat, is this shell command destructive, does this text leak a secret. CLI + MCP server.
Offline repository inspection for coding-agent instructions: scan, compare, explain, and export a local review.
CI-grade evidence and safety tools for AI agents, MCP servers, and open-source contribution workflows.
Official Python SDK for agentctl — sign, hash-chain, and govern every agent action. Zero dependencies.
Runtime governance for AI agents — budget limits, loop protection, token limits, audit trail
An honest pre-deployment reliability and security gate for AI agents.
Static security analysis for MCP servers and AI-agent workspace configurations. Taint-graph based detection of tool-description injection, capability overreach, and exfiltration paths.
Secrets detection and redaction guard plugin for AgenticStack
Compound-path and cross-release analysis of what an AI agent is permitted to do
A vendor-neutral policy-as-code toolkit for governing AI agents and agentic workflows.
CLI security scanner for AI agents: probes your HTTP endpoint (or scans traces you already have) for leaked secrets and system prompts, including the reasoning channel. Runs keyless with --demo; CI-gating exit codes.
Shared fingerprinting, comparison and baseline-sealing core for AgenTrust agent-integrity capture engines
AgentSecure SDK — LLM DLP proxy for Python
Agenttic — certify, trace and enforce AI agents. One install, one import, a signed safety grade.
Local containment runtime and firewall for AI coding agents
Official Argorix Agent Guardrails SDK for Python agents and tool execution
Contract-bounded autonomy for LLM agents: five small primitives that make an agent safe to leave alone.
Wrap your Anthropic / OpenAI Python client. Every tool call is inspected by Clavenar before it runs.
AI Agent OWASP Agentic Top 10 Compliance Scanner
Fail-closed Custodian policy guard for Claude Code, with unstrippable managed-hook enforcement
AI-security framework — build agents that are safe by construction (placeholder, under active development)
Python SDK for the ELLM Guardrail — deterministic policy firewall for AI agent tool calls
Fairness definitions and bias metrics for large language models
Gate LangChain tool calls behind a Fidacy verdict. Wraps the tool itself, so a denied payment never executes: callbacks can only observe, this can stop.
Predictive cost, policy, and runaway-session guardrails for AI coding agents
Detects toxic language in LLM-generated text using an LLM as the detection backbone. Evaluates text across seven toxicity categories: toxicity, severe toxicity, obscene, threat, insult, identity attack, and sexual explicit content.
Open, IDP-agnostic agent-identity toolkit: agent + on-behalf-of tokens via RFC 8693.
AI security lab — attack and defense testing for RAG pipelines and tool-using agents
Inter-agent security firewall for multi-agent AI systems
A domain-agnostic safety layer for AI agents: declarative policy, dry-run, audit ledger and circuit breaker between an agent's proposed actions and their execution.
Aidress agent trust and discovery tools for LangChain
AxioRank middleware for LangChain agents: govern every tool call and model turn through one Zero-Trust security gateway.
Fidacy action firewall for LangChain tool calls. Convention-named entry point: the implementation lives in fidacy-langchain, and this package re-exports it unchanged.
LangChain integration for Preloop — the open-source AI agent control plane (MCP firewall, model gateway, human approvals, budgets, and a tamper-evident audit ledger).
LangChain tools and a mandatory pre-execution gate for RelayShield's MCP registry risk and prompt-injection breach checks.
Provider-agnostic middleware for LLM safety
Runtime prompt-injection defense middleware for LLM applications
Zero-dependency scrubbers for LLM chat output — strip leaked JSON state-op fences, backend scaffold blocks, and tool-preamble meta-narration from player-facing text
Defensive pipeline for LLM apps: prompt injection, tool-call gating, output/PII guards
LLM Firewall: a safety layer between your application and any LLM API — input/output firewalling, PII redaction, jailbreak detection, and domain-specific guardrails.
Python SDK for LLMSafeSpaces API
Identifiers designed to survive being read, copied and re-emitted by language models
AIGP governance adapter for anthropic
AIGP governance adapter for Microsoft AutoGen multi-agent conversations
AIGP governance adapter for CrewAI multi-agent orchestration
AIGP governance adapter for llamaindex
AIGP governance adapter for openai
MCP-Bastion security helpers for Microsoft AutoGen / Agent Framework tool messages.
MCP-Bastion security helpers for LangGraph nodes and tool calls.
MCP-Bastion security middleware for LiteLLM (100+ LLM providers via one OpenAI-compatible API).
MCP-Bastion security middleware for Ollama local models (OpenAI-compatible API).
MCP-Bastion security helpers for OpenAI Agents SDK tools and messages.
Agent identity & governance control plane: policy-gated, credential-brokered, cryptographically audited tool execution for autonomous agents.
Python SDK for the Open Agent Specification (OAS) — decentralized identity for autonomous entities
OpenAgent SDK — identity, authentication, and Agent Capability Token verification for autonomous agents.
OpenAI Agents SDK tools and a mandatory pre-execution gate for RelayShield's MCP registry risk and prompt-injection breach checks.
Guard a custom LangGraph agent with the OpenGuardrails (OGR) protocol — a drop-in ToolNode at the agent-hook altitude, with native LangGraph interrupt() as the human-confirm gate.
Ory Agent Security for the Microsoft Agent Framework — per-tool authorization, activity logging, and identity propagation via function middleware. Built on ory-argus.
Ory Agent Security for CrewAI — per-tool authorization, activity logging, and identity propagation by wrapping tools. Built on ory-argus.
Ory Agent Security for LangChain / LangGraph — per-tool authorization, activity logging, and identity propagation via an AgentMiddleware. Built on ory-argus.
Ory Agent Security for LlamaIndex — per-tool authorization, activity logging, and identity propagation by wrapping FunctionTools. Built on ory-argus.
Ory Agent Security for the OpenAI Agents SDK — per-tool authorization, activity logging, and identity propagation via a tool-input guardrail. Built on ory-argus.
Ory Agent Security for Pydantic AI — per-tool authorization, activity logging, and identity propagation via a WrapperToolset. Built on ory-argus.
CrewAI integration for qarai-agent-guard that provides runtime security for AI agents through threat detection, policy enforcement, and content protection.
LangChain integration for qarai-agent-guard that provides runtime security for AI agents through threat detection, policy enforcement, and content protection.
RAG indirect prompt injection guard: scan retrieved documents for prompt-injection (instruction override, system/tool-prompt manipulation, exfiltration, invisible-Unicode/homoglyph smuggling) in English and Turkish, and quarantine poisoned docs before they reach the model.
🛡️ Experimental security guardrails for LangChain agent code execution — Docker sandbox isolation.
Sentinel AI — client SDK for AI agent governance & identity. Sign agent actions with Ed25519 and enforce them through the Sentinel authorization gateway. Private beta: [email protected]
CrewAI adapter for Sentinel AI agent governance — wrap any CrewAI tool so every call is authorized (and audited) by the Sentinel gateway before it runs. Private beta: [email protected]
LangChain adapter for Sentinel AI agent governance — wrap any LangChain tool so every call is authorized (and audited) by the Sentinel gateway before it runs. Private beta: [email protected]