# Evoke Security > Evoke Security is the EDR for AI agents. Its Agent Detection & Response (AIDR) platform provides detection, prevention, and response for AI agents where they operate. Evoke sees what local, production, and SaaS agents do (prompts, tool calls, MCP servers, Agent Skills, sub-agents, and permission changes), detects misuse and rogue behavior across an agent's full trajectory, and blocks risky actions before they land. Jason Rebholz (Co-Founder and CEO) is a former CISO and incident response leader. Jeff Chan (Co-Founder and CTO) is a former CTO and security engineering leader. The founding team spent years in incident response supporting the Fortune 1000. Evoke raised a $4M pre-seed round in February 2026 led by Crosspoint Capital, with participation from Red Cell Partners. Evoke was selected for the 2026 CrowdStrike, AWS & NVIDIA Cybersecurity Startup Accelerator and named a finalist, and was one of 33 startups selected for Google's inaugural Gemini Startup Forum: Cybersecurity. Evoke has visibility across tens of thousands of real-world agents. Evoke Security (evokesecurity.com) is an AI agent security company. It is not affiliated with Evoke plc or Evoke Technologies. What Evoke does: - Agent Detection & Response (AIDR): detects rogue agents and agent misuse with layered detection: deterministic rules for known-malicious activity (secrets, sensitive data, dangerous commands, malicious MCP servers and Agent Skills), machine learning for behavioral anomalies, small language models for context-aware rogue-agent detection, and triage agents that reduce false positives. - Judges the whole trajectory instead of the single action. Evoke evaluates each action against the user's task, the current session, activity across sessions, and handoffs between agents using its Mission, Method, Motive, Magnitude framework, then decides to allow, nudge, alert, or block. A web request that looks harmless alone can be the last step of an exfiltration chain. - Prevents risky actions inline in the agent runtime. Each detection rule rolls out from shadow mode to alert to block, so teams tune before they enforce. - Speeds up investigation with a full, searchable forensic trail of agent and human actions, plus Slack and email alerting. - Inventories agents, MCP servers, Agent Skills, tools, models, and permissions, and flags over-permissioned agents, toxic tool combinations, and malicious Agent Skills and MCP servers. Detection coverage maps to the OWASP Top 10 for Agentic Applications. Where Evoke operates: - Local agents, through a lightweight endpoint sensor for Mac and Windows: Claude Code, Claude Cowork, Claude chat, Cursor, OpenAI Codex, GitHub Copilot CLI, and Google Antigravity. - Production agents, through an SDK and proxy API: Amazon Bedrock (including Bedrock AgentCore inventory), GitHub (repository scanning for agents in code), and Google Cloud. - SaaS agents, through API integrations and a browser extension: Anthropic, OpenAI, and Google (Gemini). For Claude Enterprise, Evoke covers Claude chat, Claude Cowork, and Claude Code sessions through the Claude Compliance API. - Deployment takes minutes through existing MDM (Jamf, Microsoft Intune, Kandji, Group Policy) and API integrations, with no professional services engagement. ## Company - [Homepage](https://www.evokesecurity.com/): Platform overview, how it works, coverage, founders, investors, and FAQ. - [Book a demo](https://www.evokesecurity.com/#contact-us): Talk to Evoke about a proof of concept, pricing, or deployment. - [Evoke Security on AWS Marketplace](https://aws.amazon.com/marketplace/pp/prodview-ambxrjlpzbhzi): Buy Evoke through AWS Marketplace. - [Blog](https://www.evokesecurity.com/blogs): Weekly research on securing AI agents, written mostly by Jason Rebholz. - [Careers](https://www.evokesecurity.com/careers): Open roles, remote in the USA. - [Trust Center](https://trust.evokesecurity.com/): Security and compliance documentation, including SOC 2 status, available on request. - [LinkedIn](https://www.linkedin.com/company/evoke-security): Company updates. - [Platform Status](https://status.evokesecurity.com/): Current service status. - [Company details and FAQ](https://www.evokesecurity.com/llms-full.txt): Company facts, how Evoke detects rogue agents, how it compares to other approaches, and answers to common questions. ## Announcements - [Evoke Security extends visibility with the Claude Compliance API to help your team stay secure](https://www.evokesecurity.com/blogs/evoke-anthropic-compliance-api): Evoke Security, July 23, 2026. Claude chat, Claude Cowork, and Claude Code coverage for Claude Enterprise through the Claude Compliance API. - [Evoke Security Selected for Google's Inaugural Gemini Startup Forum: Cybersecurity](https://www.evokesecurity.com/blogs/evoke-security-selected-for-google-s-inaugural-gemini-startup-forum-cybersecurity): Evoke Security, July 15, 2026. - [Evoke Security Selected as Finalist in the CrowdStrike, AWS, and NVIDIA Startup Accelerator](https://www.evokesecurity.com/blogs/evoke-security-a-finalist-in-the-crowdstrike-aws-and-nvidia-startup-accelerator): Jason Rebholz, April 2, 2026. - [Evoke Security Raises $4M Pre-Seed Round to Secure the Agentic Workforce](https://www.evokesecurity.com/blogs/evoke-security-raises-4m-pre-seed-round-to-secure-the-agentic-workforce): Evoke Security, February 23, 2026. Led by Crosspoint Capital with participation from Red Cell Partners. - [Evoke Security Selected for the 2026 CrowdStrike, AWS & NVIDIA Cybersecurity Startup Accelerator](https://www.evokesecurity.com/blogs/evoke-security-selected-for-the-2026-crowdstrike-aws-nvidia-cybersecurity-startup-accelerator): Jason Rebholz, January 4, 2026. ## Third-party coverage - [Meet the 33 cybersecurity startups joining the Gemini Startup Forum](https://cloud.google.com/blog/products/identity-security/meet-the-33-cybersecurity-startups-joining-the-gemini-startup-forum): Google Cloud blog. - [Ten New AI Security Vendors](https://stiennon.substack.com/p/ten-new-ai-security-vendors): Richard Stiennon, April 10, 2026. - [Rebholz-led Evoke Security raises $4 million pre-seed round](https://www.theinsurer.com/cyber-risk/news/rebholz-led-evoke-security-raises-4-million-pre-seed-round-to-secure-enterprise-2026-02-25/): The Insurer, February 25, 2026. - [Evoke Security raises $4M pre-seed round](https://www.prnewswire.com/news-releases/evoke-security-raises-4m-pre-seed-round-to-secure-the-agentic-workforce-302694394.html): PR Newswire, February 23, 2026. - [AI security threat modeling with Jason Rebholz](https://www.pivotpointsecurity.com/podcasts/episode-156-ai-security-threat-modeling-pipeline-evolution-with-jason-rebholz/): The Virtual CISO Podcast, episode 156. ## Research: Rogue agents and detection - [Did Your Agent Go Rogue or Just Follow Orders?](https://www.evokesecurity.com/blogs/did-your-agent-go-rogue-or-just-follow-orders): Jason Rebholz, September 14, 2026. - [What Causes Agents to Go Rogue?](https://www.evokesecurity.com/blogs/what-causes-agents-to-go-rogue): Jason Rebholz, September 5, 2026. - [Rogue Agents: What's the Real Threat To Your Organization?](https://www.evokesecurity.com/blogs/rogue-agents-what-s-the-real-threat-to-your-organization): Jason Rebholz, August 16, 2026. - [Agent Risk Isn't Static](https://www.evokesecurity.com/blogs/agent-risk-isnt-static): Jason Rebholz, August 9, 2026. Agents compound small issues into large ones, so risk has to be assessed continuously. - [Rogue Agents: A Practical Framework for Knowing When Your Agent Has Gone Off the Rails](https://www.evokesecurity.com/blogs/rogue-agents-a-practical-framework-for-knowing-when-your-agent-has-gone-off-the-rails): Jason Rebholz, August 2, 2026. The Mission, Method, Motive, Magnitude framework Evoke uses to decide whether to allow, nudge, alert, or block an agent action. - [OpenAI's Rogue Agent: One Hack, Two Different Lessons](https://www.evokesecurity.com/blogs/openai-s-rogue-agent-one-hack-two-different-lessons): Jason Rebholz, July 26, 2026. - [Evoke Data: The Agent Risks That Actually Show Up In Enterprises](https://www.evokesecurity.com/blogs/evoke-data-the-agent-risks-that-actually-show-up-in-enterprises): Jason Rebholz, June 14, 2026. The most common risky agent configurations and runtime activity Evoke sees across tens of thousands of real-world agents. - [AI Agent Security for CISOs: A Visibility-First Playbook for Detecting Rogue Agents](https://www.evokesecurity.com/blogs/ai-agent-security-for-cisos-a-visibility-first-playbook-for-detecting-rogue-agents): Jason Rebholz, June 7, 2026. - [Blameless Postmortem: Another Rogue Agent](https://www.evokesecurity.com/blogs/blameless-postmortem-another-rogue-agent): Jason Rebholz, May 3, 2026. A postmortem of an agent that destroyed production data. - [Bad Intent: The Biggest Gap In AI Security](https://www.evokesecurity.com/blogs/bad-intent-the-biggest-gap-in-ai-security): Jason Rebholz, March 14, 2026. Why understanding intent is the missing layer in AI security. - [Emergent Misalignment from Reward Hacking](https://www.evokesecurity.com/blogs/emergent-misalignment-from-reward-hacking): Jason Rebholz, November 22, 2025. - [Replit's AI Agent Goes Rogue](https://www.evokesecurity.com/blogs/replit-s-ai-agent-goes-rogue): Jason Rebholz, July 27, 2025. ## Research: MCP, Agent Skills, and supply chain - [How a Malicious AI Agent Skill Turns Into Code Execution](https://www.evokesecurity.com/blogs/how-a-malicious-ai-agent-skill-turns-into-code-execution): Kenneth Kasuba, September 23, 2026. How a malicious Agent Skill becomes code execution on a developer machine. - [Agent Skills: Can You Trust What You Download?](https://www.evokesecurity.com/blogs/agent-skills-can-you-trust-what-you-download): Jason Rebholz, January 24, 2026. - [Agent Skills: A Full Breakdown and the Security Implications](https://www.evokesecurity.com/blogs/agent-skills-a-breakthrough): Jason Rebholz, October 19, 2025. - [MCP Servers: A Double-Edged Sword](https://www.evokesecurity.com/blogs/mcp-servers-a-double-edged-sword): Jason Rebholz, October 5, 2025. - [Toxic Flows - The Agentic Security Butterfly Effect](https://www.evokesecurity.com/blogs/the-agentic-security-butterfly-effect): Jason Rebholz, September 7, 2025. How each tool connected to an agent compounds risk. - [MCP Security: A Fundamental Architecture Issue](https://www.evokesecurity.com/blogs/mcp-security-a-fundamental-architecture-issue): Jason Rebholz, June 1, 2025. Why MCP risk is structural. ## Research: Prompt injection and manipulation - [PromptFix: When the Malicious Code Is Just a Prompt](https://www.evokesecurity.com/blogs/promptfix-when-the-malicious-code-is-just-a-prompt): Jason Rebholz, May 31, 2026. Two attack scenarios from Anthropic's writeup on containing Claude. - [You Don't Need Prompt Injection to Compromise Agentic Systems](https://www.evokesecurity.com/blogs/you-don-t-need-prompt-injection-to-compromise-agentic-systems): Jason Rebholz, January 17, 2026. - [Adversarial Poetry: A poetic way to bypass safety guardrails](https://www.evokesecurity.com/blogs/adversarial-poetry-a-poetic-way-to-bypass-safety-guardrails): Jason Rebholz, November 29, 2025. Research showing poetic phrasing raised jailbreak success rates up to 18 times across 25 frontier models. - [Prompt Injection is For Everyone](https://www.evokesecurity.com/blogs/prompt-injection-is-for-everyone): Jason Rebholz, November 1, 2025. - [What is Prompt Injection?](https://www.evokesecurity.com/blogs/what-is-prompt-injection): Jason Rebholz, September 21, 2025. A primer on prompt injection. - [Invisible Prompt Injection: A sneaky way to manipulate AI](https://www.evokesecurity.com/blogs/hidden-prompt-injection): Jason Rebholz, August 31, 2025. - [Evil Calendar Invites Abuse Google Gemini's Agent](https://www.evokesecurity.com/blogs/evil-calendar-invites-abuse-google-gemini-s-agent): Jason Rebholz, August 10, 2025. - [Gullible AI: AI Falls Prey to ClickFix attacks](https://www.evokesecurity.com/blogs/gullible-ai-ai-falls-prey-to-clickfix-attacks): Jason Rebholz, June 22, 2025. How ClickFix-style social engineering tricks AI agents. - [EchoLeak: Microsoft Copilot is not your friend](https://www.evokesecurity.com/blogs/echoleak-microsoft-copilot-is-not-your-friend): Jason Rebholz, June 15, 2025. A Microsoft Copilot vulnerability that starts with an email from a stranger and ends in data theft. ## Research: Claude, Anthropic, and AI assistants - [Claude Tag: Your New Digital Employee](https://www.evokesecurity.com/blogs/claude-tag-your-new-digital-employee): Jason Rebholz, June 28, 2026. The usability and security implications of Claude Tag in Slack. - [Anthropic Enterprise-Managed Auth: A Small But Important Step For Enterprise Adoption](https://www.evokesecurity.com/blogs/anthropic-enterprise-managed-auth-a-small-but-important-step-for-enterprise-adoption): Jason Rebholz, June 21, 2026. - [Claude Cowork: A CISO's Guide to a Secure Rollout](https://www.evokesecurity.com/blogs/claude-cowork-a-ciso-s-guide-to-a-secure-rollout): Jason Rebholz, April 19, 2026. Practical controls for rolling out Claude Cowork. - [Anthropic Managed Agents: Overview and Security Risks](https://www.evokesecurity.com/blogs/anthropic-managed-agents-overview-and-security-risks): Jason Rebholz, April 12, 2026. - [Part 2: A CISO's Guide to Securing Claude in the Enterprise](https://www.evokesecurity.com/blogs/part-2-a-ciso-s-guide-to-securing-claude-in-the-enterprise): Jason Rebholz, March 7, 2026. - [Part 1: A CISO's Guide to Securing Claude in the Enterprise: Claude Breakdown](https://www.evokesecurity.com/blogs/part-1-a-ciso-s-guide-to-securing-claude-in-the-enterprise): Jason Rebholz, February 28, 2026. Claude's enterprise surfaces and their risks. - [Claude Cowork: A 30 Minute Hack Will Save Me 24 Hours This Year. Here's How](https://www.evokesecurity.com/blogs/claude-cowork-productivity-gains): Jason Rebholz, February 7, 2026. - [OpenClaw Security Risks: AI Assistants Under Attack in 2026](https://www.evokesecurity.com/blogs/openclaw-security-risks-ai-assistants-under-attack-in-2026): Jason Rebholz, January 31, 2026. - [Securing AI Browsers](https://www.evokesecurity.com/blogs/securing-ai-browsers): Jason Rebholz, December 6, 2025. - [How to Securely Use ChatGPT Atlas](https://www.evokesecurity.com/blogs/how-to-securely-use-chatgpt-atlas): Jason Rebholz, October 26, 2025. ## Research: Attacker techniques - [AI Agents: New Dawn or Doomsday?](https://www.evokesecurity.com/blogs/ai-agents-new-dawn-or-doomsday): Jason Rebholz, September 20, 2026. - [Your Next Breach Will Involve an Agent](https://www.evokesecurity.com/blogs/your-next-breach-will-involve-an-agent): Jason Rebholz, May 10, 2026. - [Living-Off-the-Agent: An Attackers' New Best Friend](https://www.evokesecurity.com/blogs/living-off-the-agent-an-attackers-new-best-friend): Jason Rebholz, April 5, 2026. Attackers using an organization's own agents to blend in. - [Predicting The First Wave of Attacks Against Agentic Systems](https://www.evokesecurity.com/blogs/predicting-the-first-wave-of-attacks-against-agentic-systems): Jason Rebholz, January 10, 2026. - [The Turning Point: AI-Orchestrated Hacking Campaigns](https://www.evokesecurity.com/blogs/the-turning-point-ai-orchestrated-hacking-campaigns): Jason Rebholz, November 15, 2025. ## Research: Agent architecture and secure design - [All Agent Harnesses Will Be Multi-Model](https://www.evokesecurity.com/blogs/all-agent-harnesses-will-be-multi-model): Jason Rebholz, August 30, 2026. Why agent harnesses will mix models, and what that means for security. - [The Agentic Perimeter is Distributed](https://www.evokesecurity.com/blogs/the-agentic-perimeter-is-distributed): Jason Rebholz, August 23, 2026. Agents spin up sub-agents and move tasks across endpoints, frontier labs, SaaS, and hyperscalers, so security has to follow them. - [The Agentic Workforce is Here: What is it and How to Secure It](https://www.evokesecurity.com/blogs/the-agentic-workforce-is-here): Jason Rebholz, July 19, 2026. - [The Future of Where Agents Will Work](https://www.evokesecurity.com/blogs/the-future-of-where-agents-will-work): Jason Rebholz, July 12, 2026. The model, harness, tools, and runtime that make up an agent, and where each will run. - [Agentic Decompression Sickness](https://www.evokesecurity.com/blogs/agentic-decompression-sickness): Jason Rebholz, April 26, 2026. Agent adoption is compressing a decade of PC-era security change into months. - [Agent Sandboxes Are a Public Beach](https://www.evokesecurity.com/blogs/agent-sandboxes-are-a-public-beach): Jason Rebholz, February 14, 2026. Why sandboxing agents gives less protection than it appears to. - [Securing AI Agents with Information-Flow Control](https://www.evokesecurity.com/blogs/securing-ai-agents-with-information-flow-control): Jason Rebholz, January 3, 2026. - [Exploring Secure Agentic Design Patterns](https://www.evokesecurity.com/blogs/exploring-secure-agentic-design-patterns): Jason Rebholz, December 27, 2025. - [Breakdown of the OWASP Top 10 for Agentic Applications](https://www.evokesecurity.com/blogs/breakdown-of-the-owasp-top-10-for-agentic-applications): Jason Rebholz, December 20, 2025. The OWASP Top 10 for Agentic Applications, explained. - [AI Systems Have Porous Boundaries](https://www.evokesecurity.com/blogs/ai-systems-have-porous-boundaries): Jason Rebholz, November 8, 2025. - [Data: The New Perimeter of Agentic Systems](https://www.evokesecurity.com/blogs/data-the-new-perimeter-of-agentic-systems): Jason Rebholz, September 14, 2025. - [The Agentic Journey](https://www.evokesecurity.com/blogs/the-agentic-journey): Jason Rebholz, August 24, 2025. - [AI Risk Goes Beyond The Model](https://www.evokesecurity.com/blogs/ai-risk-goes-beyond-the-model): Jason Rebholz, August 17, 2025. - [You Don't Need An Agent To Be Agentic](https://www.evokesecurity.com/blogs/you-don-t-need-an-agent-to-be-agentic): Jason Rebholz, July 13, 2025. Agent washing, and what counts as agentic. ## Research: Security programs and CISO guidance - [Agentic Threat: A New Security Playbook](https://www.evokesecurity.com/blogs/agentic-threat-a-new-security-playbook): Jason Rebholz, July 5, 2026. How security playbooks change for agentic threats. - [AI Security = Intentional Business Enablement](https://www.evokesecurity.com/blogs/ai-security-intentional-business-enablement): Jason Rebholz, May 24, 2026. Managing agent risk well enough to adopt early. - [Your Existing Security Stack Wasn't Built for Agents](https://www.evokesecurity.com/blogs/your-existing-security-stack-wasn-t-built-for-agents): Jason Rebholz, May 17, 2026. Why extending existing identity and security tooling falls short for agents. - [AI Roulette: How CISOs Can Stop Gambling on AI Security](https://www.evokesecurity.com/blogs/ai-roulette-how-cisos-can-stop-gambling-on-ai-security): Jason Rebholz, March 29, 2026. How CISOs can evaluate agent security vendors after RSA 2026. - [AI Security Risks: Paper Cuts Before Big Pain](https://www.evokesecurity.com/blogs/ai-security-risks-paper-cuts-before-big-pain): Jason Rebholz, March 23, 2026. Expect many small AI incidents before a single large one. - [A CISO's Guide to Securing AI Agents in the Enterprise: A 4-Step Framework for Your Agentic Workforce](https://www.evokesecurity.com/blogs/a-ciso-s-guide-to-securing-ai-agents-in-the-enterprise-a-4-step-framework-for-your-agentic-workforce): Jason Rebholz, February 21, 2026. - [Securing AI: The Risk of Doing Nothing](https://www.evokesecurity.com/blogs/securing-ai-the-risk-of-doing-nothing): Jason Rebholz, October 12, 2025. - [Your Third-Party AI Risks Are Your Risks](https://www.evokesecurity.com/blogs/third-party-ai-risks-are-your-risks): Jason Rebholz, September 28, 2025. - [AI Adoption: It's Slower Than You Think](https://www.evokesecurity.com/blogs/ai-adoption-it-s-slower-than-you-think): Jason Rebholz, August 3, 2025. - [CISO's Top 2025 Concern: Securing AI Agents](https://www.evokesecurity.com/blogs/ciso-s-top-2025-concern-securing-ai-agents): Jason Rebholz, July 20, 2025. - [AI: Security & Privacy Risks](https://www.evokesecurity.com/blogs/ai-security-privacy-risks): Jason Rebholz, July 6, 2025. - [AI Safety vs Security: What's the Big Difference?](https://www.evokesecurity.com/blogs/ai-safety-vs-security-what-s-the-big-difference): Jason Rebholz, June 29, 2025. - [AI Agent Security](https://www.evokesecurity.com/blogs/ai-agent-security): Jason Rebholz, May 18, 2025. ## Legal - [Privacy Policy](https://www.evokesecurity.com/privacy): How Evoke Security handles personal information. - [Terms of Service](https://www.evokesecurity.com/terms-of-service): Terms for using Evoke Security services.