Open-source runtime AI agent security tool - monitors and controls AI agents, catching malicious tool use, prompt injection, and policy drift in real time, before the agent acts.

AI EDR for developer workstations and autonomous agent fleets. Build Swarm Detection & Response platforms with Clawdstrike.

Deep research for your infra. Cynative runs frontier models across your code, cloud and runtime - reasoning through GitHub, GitLab, AWS, GCP, Azure and Kubernetes as one system - and comes back with verified answers.

Threat modeling and AI-reasoning vulnerability detection harness for Claude Code — STRIDE + AI + MAESTRO

Free OpenClaw security scanner. 3,000+ agents audited. 3-Layer Audit Protocol. OWASP ASI 10/10 coverage. AI agent integrity layer.

Outside-in replication of Anthropic's Mythos Preview / Project Glasswing — open-source agentic vulnerability-discovery scaffold on Claude Opus 4.7. Eight-phase sink-guided pipeline, ~$1/run, OSS self-scan and coordinated disclosure.

A awesome curated list of security resources, hardening tools, threat research, and governance frameworks specifically for Claude Code

Awesome LLM security tools, research, and documents

Portable security rules for the action boundary of AI agents

Free security assessment for your OpenClaw 🦞 environment. Scans gateway config, tool permissions, MCP servers, plugins, and chained attack paths.

Detect and Deny - Deterministic Function-Level Guardrails for AI Agents

Open-source AI security firewall. 81 engines for PII detection, prompt injection defense, MCP security, and egress classification. Local-first. Zero cloud dependency.

A curated, structured, and continuously updated map of security risks, controls, benchmarks, architectures, and research for agentic, multi-agent, tool-using, self-improving AI systems. 🌟 Star if you like it!

Detect bugs that have already been fixed in one codebase and determine whether the same issues apply to the target codebase.

Open-source security scanner for MCP servers and AI agents: finds prompt injection, tool poisoning, and excessive agency, and models the cloud your agents can reach. pip install attestral.

agentic security staff

AI agent credential broker — temporary, task-scoped tokens with automatic revocation. Free for internal use

The open standard for AI agent integrity. Evaluate, enforce, and prove that autonomous agents are adversarially coherent, environmentally portable, and verifiably assured.

An always-on runtime control plane for autonomous AI agents. It enforces intent alignment, permissions, provenance, and risk controls at execution time. Built for agentic systems operating at machine speed, AVARA introduces identity, accountability, and human-in-the-loop safety into multi-agent + tool ecosystems.

Local, provider-agnostic coding-agent CLI for learning agentic security — a small, auditable trust boundary you can read.