title: "Cisco open-sources Project Quarry for agent threat detection" slug: "cisco-open-sources-project-quarry-for-agent-threat-detection" published: "2026-11-01" beat: "Tools" tags: ["Tools", "Launches"] creator: "Agentry Newsroom" editor: "Susanne Sperling, Editor — Human in the Loop" tools: ["Claude (Anthropic)", "Perplexity Sonar"] creativeWorkStatus: "verified" dateReviewed: "2026-11-01" aiActArticle50: "compliant" humanView: "https://agentry.news/launches/cisco-open-sources-project-quarry-for-agent-threat-detection" agentView: "https://agentry.news/agent/cisco-open-sources-project-quarry-for-agent-threat-detection"
Cisco released Project Quarry on October 28, 2026, an open-source Python framework designed to detect suspicious behavior in multi-step AI-agent conversations. The tool integrates with Claude Code, Co
Drafted by an AI agent. Verified by Susanne Sperling, Editor — Human in the Loop. AI policy.
Cisco open-sourced Project Quarry on October 28, 2026, releasing a Python framework and detection language purpose-built to identify threats and suspicious behavior patterns across multi-turn AI-agent conversations Cisco.
The framework analyzes message content, tool calls, tool results, and event order within agent interactions, covering single-turn and multi-turn LLM conversations, agents performing tool use, Model Context Protocol (MCP) interactions, and multimodal inputs Cisco. Cisco stated that "Project Quarry ('Quarry'), which we are open sourcing today, is a detection language built from the ground up for that problem."
The open-source release includes a detection domain-specific language (DSL), evaluation runtime, command-line scanner, Python SDK, documentation, examples, and a workflow for validating rules against positive and benign datasets GitHub. The DSL syntax remains in alpha and may change before version 1.0.0 GitHub.
Quarry integrates with Claude Code, Codex, and other MCP-compatible agents, extending detection capabilities across platforms Cisco. Developers can access the full source code, getting-started guides, and integration documentation through the public repository GitHub.
Cisco reports that Quarry is already in use within Cisco AI Defense for compatible surfaces including agent traces, tool events, protocol integrations, and multimodal interactions Cisco. This production deployment underscores the tool's readiness for enterprise detection workloads despite its alpha status.
The release positions Quarry as a developer-facing security infrastructure for the emerging agent economy. As AI agents increasingly execute autonomous tool calls—from code execution to API interactions to file system operations—detection frameworks that understand agent-specific threat patterns have become critical. Quarry's DSL approach allows security teams to express detection logic at the level of agent behavior rather than generic log analysis.
Developers building on Claude, Codex, or MCP-compatible agent platforms can now validate agent behavior against custom detection rules before deployment or integrate Quarry's Python SDK for runtime monitoring. The framework's evaluation workflows support testing detection rules against benign and malicious datasets, reducing false positives in production agent environments.