AI guardrails & agent security
3,625 agents · most starred first
AI guardrails & agent security, page 32 of 37
Enterprise-grade LLM security guardrail library for protecting input prompts and output responses
Citation-verified effective-access analysis for AI agent fleets.
Local-first AI agent inventory and risk scanner: finds agent candidates with evidence, maps capabilities and integrations, and gates newly introduced high-confidence critical risks.
GEPA reflective prompt evolution optimizer for superred agent targets
PoisonedRAG knowledge-corruption optimizer for superred
Deny-by-default command safety gate and text-ReAct routing agent for small LLMs
Deterministic governance filter for Semantic Kernel — function-call authorization, PII scanning, cost budgets, and structured audit trail.
LangChain request-signing adapter for the VAID standard: sign every outbound agent HTTP request with a VAID proof-of-possession via a thin httpx.Auth seam.
Hardened single-node registry service for Agent Auth identities
Python SDK for agent identity publishing, signed messaging, and registry-based verification
Sign OpenAI agent tool calls with Vouch Credentials.
Drop-in payload firewall SDK for agent services, with APA v0.1 protection proofs.
Client SDK for Warden — build and sign canonical Warden delegation tokens, and route agent frameworks through the Warden MCP proxy.
Phronesis review: compare features, pricing, use cases, access model, and alternatives for this Model Context Protocol (MCP) agent in 2026. A neutral decision-assurance layer for autonomous agent actions.
Evidence-based scoped AI agent reputation with W3C Verifiable Credentials 2.0 adapters.
Consent-based local observer for Phigon AI agent supervision.
Agent Approve extension for Pi - approve or deny AI agent tool calls from your iPhone and Apple Watch
TypeScript SDK for Atbash — the safety layer that evaluates AI agent actions against operator-defined policies before execution.
Versioned provider-neutral policy lifecycle, simulation, and AuthZEN adapter for AI agent actions.
Ory Argus: the core API for building authentication, authorization, and audit into AI agent harness plugins, extensions, and custom integrations
Dependency-free static analyzer that audits LLM agent tool/function definitions (OpenAI, Anthropic, MCP) for excessive agency, code-execution & SSRF sinks, secret access, and data-exfiltration surface — mapped to the OWASP 2025 LLM Top-10. Runs in CI.
High-risk-only approval hook with LLM risk analysis, behavior detection, protected-path interception, and decision memory for oh-my-pi (OMP) and pi-agent.
Defend an agent against hidden-content injection: strip payload-capable invisible Unicode and ANSI, splice out human-invisible HTML, and flag data-exfil URLs in untrusted text before any model sees it.
NexusGenesis Agent Autonomous Key Security — PQC (Dilithium2) keys, AES-256-GCM encryption, three-tier key derivation, and human takeover. Private keys never leave the agent/browser.
ESLint plugin for Vercel AI SDK security — detects prompt injection, system-prompt leaks, hardcoded API keys, and unvalidated model output in generateText and streamText.
Command-line security tool that verifies what AI coding agents can actually read, write, execute and reach using synthetic canaries.
Self-describing encrypted env files for AI-assisted development
AgentAvow (formerly agentgraph-agt) trust provider for Microsoft Agent Governance Toolkit
AgentAvow (formerly agentgraph-pydantic) trust verification middleware for PydanticAI agents
Python SDK and CLI for the AgentAvow trust and identity platform (formerly agentgraph-sdk)
Agent-neutral policy-engineering framework
Shared invoke-boundary core for agent-governance-demo: host adapters, entitlement/policy resolution, the governance envelope, and the registry Postgres schema. Published to PyPI via OIDC Trusted Publishing (.github/workflows/publish.yml) and installed into registry-api/workflows images.
Open-source security for AI coding agents: record, detect and block at the tool boundary, with a verifiable audit trail and SIEM forwarding
Python SDK for Agentomy AI Governance Platform
Deterministic security scanner for AI agent skills + Trusted Distribution CLI. Local, zero-dependency, executes nothing. Facts, not verdicts.
TollWarden payment security for CrewAI agents: scan x402 payments before settling, auto-tag provenance for prompt-injection detection via a tool-call hook, and guard payment tools so blocked payments can never execute.
Deterministic Adversarial Simulation (DAS-LLM) Framework for LLM Security Invariants
Self-certifying agent identity: keys, addresses, did:amp, signed cards, owners, and delegation chains.
TollWarden payment security for LangChain/LangGraph agents: scan x402 payments before settling, auto-tag provenance for prompt-injection detection, guard payment tools so blocked payments can never execute.
Sub-20ms prompt injection and jailbreak protection for LlamaIndex query engines.
Practical, FOSS-backed security gates for LLM systems: model-file supply-chain gate, runtime I/O firewall proxy, and model resilience scoring.
AI-security auditor — score any AI endpoint against Responsible-AI metrics, plus a 44-control safety register (early/experimental)
Audit trail, authorization, and anomaly detection for AI agents
Q SDK — Instrument any AI agent with security governance
COSE Sign1 proof-of-possession helpers for Swarmauri
Python client SDK for Vexa AgentWall AI security gateway & firewall
ProofChain review: compare features, pricing, use cases, access model, and alternatives for this ai-security agent in 2026. Verifiable AI agent compliance with soulbound identity and tamper-evident audit trails on X1.
ProofChain KYA review: compare features, pricing, use cases, access model, and alternatives for this ai-security agent in 2026. EU AI Act Article 50 compliance audit agent with tamper-proof cryptographic seals.
ClaudeWatermarks review: compare features, pricing, use cases, access model, and alternatives for this AI Detection agent in 2026. Inspect hidden text artifacts locally in your browser.
Secure Agent Control Plane & Passport Vault — identity-aware reverse proxy for AI agent credentials.
Agent runtime rule engine: tool ACL -> dangerous pattern -> constitution immutability -> identity continuity
Local-first safety scanner and incident report generator for AI coding agent workspaces.
A pre-execution, deterministic policy decision point for tool-using AI agents. Evaluate every proposed tool call against declarative policy and return allow / audit / escalate / deny.
An extensible, security-aware AI agent framework with Skills and Plugins.
Zero-configuration auto-instrumentation and guardrails for multi-agent LLM frameworks.
Provenance-aware runtime security primitives for autonomous AI agents
Call approved SaaS platforms from any Python agent. Your agent never holds the downstream credential.
AgentVisa credential presentation gateway
Platform-side verification library for AgentVisa v3 authentication
LangChain integration for 1Claw — HSM-backed secrets, signing, memory, and automations for AI agents
AgentLock authorization middleware for MCP (Model Context Protocol) servers.
Signed, tamper-evident receipts for A2A and MCP agent interactions (triple-entry accounting)
AumOS Agent SDK — unified Python SDK that gives coding agents AumOS security primitives as first-class objects
Adds a company-policy check before consequential AI agent actions run.
Official TypeScript/JavaScript SDK for GATRA Zero-Trust AI Agent Security Engine
Secure WebMCP registration, policy, audit, and test adapters for browser-native AI agent actions.
Security gate for AI agent skill installers
n8n community node for Radware Agentic AI Protection in-path enforcement
Cross-platform decision evidence and reconstruction for regulated AI agents
A policy enforcement point for agent spending: blocks transactions that violate a declarative policy and writes a tamper-evident, signed audit log of every decision.
Agent control surface: vault-backed rollback, typed hooks, circuit breaker, audit log
Execution integrity and trajectory authenticity governance middleware for autonomous AI agents
Agent Security Evaluation Kit — score agent code and refuse unsafe tools, with receipts, AST checks, and live-trace adapters.
Git-backed registry and guardrails for shared agent skills.
Deterministic policy engine for LLM-generated SQL: SELECT-only, PII column denylist, table allowlist, cost cap. Multi-dialect via sqlglot (BigQuery, Snowflake, Postgres, Trino, …).
Transaction-safe policy, approval, and execution gateway for AI-agent actions
Firewall for AI coding agents. Intercepts Bash/Read/Write/Edit, blocks outbound HTTP, asks humans via Slack/Telegram for risky ops. Works with Claude Code, Cursor, Continue.dev, Aider, and GitHub Actions.
Open-source defensive harness and learning lab for reproducing and measuring agentic AI failure modes with portable traces, attack graphs, scorecards, and data-boundary tests.
AgentPerimeter — runtime governance for AI agents with CIS-region compliance (FSTEC 117, GOST R 57580, KZ AI Law). Name reservation; product at agentperimeter.ru
Cryptographically signed, tamper-evident audit logs for autonomous agents
Build TRACE Trust Records from evidence another system produced, without fabricating what is not there
A tool-agnostic policy layer that intercepts and blocks destructive commands proposed by autonomous AI coding agents, with full audit logging.
Experimental fail-closed pre-dispatch boundary for AI agent tools.
High-speed deterministic scanning for LLM inputs and outputs
GenAI model provenance and jurisdiction assurance harness
Runtime security proxy for MCP agent tool calls
Governance for AI agents: deny-by-default tool gating, L0-L4 guardrails, human approvals and a full audit chain — via the agent-governance sidecar.
Vendor-neutral Governed Contracts for verifiable agent authority, constraints, conformance and consequence.
Reference validator and applicator for the Open Agent Profile (OAP) specification
Python interoperability utilities for OpenAgentFence security traces (early development)
Open-source in-process governance for AI agent frameworks: Cedar-governed model/tool calls via build_middleware()/GovernedAgent, plus the Cedar engine, request/decision shapes, PEP<->control-plane protocol client, and Ed25519 PEP identity they run on.
Runtime enforcement library for autonomous AI agent systems: constraint enforcement, auditable decision traces, defense-in-depth safety controls, and fail-closed network resilience.
AgentVigil/WebSentinel indirect prompt-injection optimizer for superred agent targets
Environmental Injection Attack optimizer for SuperRed web agents
Open-source agent security framework. Detects and defends against AI Agent Traps - content injection, embedded jailbreaks, RAG poisoning, data exfiltration, and more.
Local tool-call failsafe for DSH: blocks a fixed list of high-risk shell patterns and credential-file reads before execution, with a per-session hash-linked JSONL audit log, and an optional external Gateway for centralized policy.
The only agent execution layer that enforces governance before consequences land. x402-native paid execution, financial control plane, immutable receipts, output verification, agent reputation, memory continuity, and Attestify Trust (no-wallet agent ident
Kernel-level Zero-Trust Security SDK for AI Agents and LLMs via Agent-eBPF
LangChain and LangGraph deterministic invariant tool guard
ADAPT-Agent: Adversarial Defense & Policy Training for LLM Agents - A comprehensive library for LLM agent optimization and security