Manifold adds browser support for AI agent governance

Manifold Security on September 16, 2026 launched browser support for its AI governance platform, extending monitoring from endpoint activity to AI assistants and browser-driving agents in managed Chromium browsers on macOS, Windows and Linux. The update gives security teams a single control plane to trace what AI agents prompt, access, and do across browser and endpoint runtime.

Why it matters: - AI agents are moving from chat windows into browsers where they can search, act in authenticated sessions and interact with external systems. - Security teams need visibility into agent behavior, not just employee use of AI tools. - Manifold’s new browser coverage is designed to show what an agent actually did across browser and endpoint runtime from one control plane.

What happened: - Manifold Security launched browser support for its AI governance platform on September 16, 2026. - The rollout extends endpoint coverage to AI activity inside managed Chromium-based browsers on macOS, Windows and Linux. - The platform traces an AI agent from the initial prompt through tool calls to the final outcome. - The browser capability has been running in customer deployments for several months and is available now.

The details: - The browser support covers two categories of AI use. - AI assistants in browser tabs include claude.ai, Claude Code on the web, Claude Cowork, ChatGPT, Gemini, Cursor and Perplexity. - Browser-driving agents include Claude in Chrome, Replit, ChatGPT and Gemini agents. - Manifold says its platform is the first capable of monitoring Claude in Chrome. - For AI assistants in browser tabs, Manifold can block a denied prompt before it leaves the browser. - The same controls can prevent a denied AI site from loading and enforce corporate account rules when required. - The platform can flag and record policy breaches in real time, with violations tied to the user, device and conversation. - Manifold can detect activity based on what the AI did, including tool calls, where data went and combinations of capabilities. - The platform can include MCP servers, skills and plugins connected to browser AI in the same AI bill of materials as endpoint activity. - For browser-driving agents, Manifold applies the same policies and detections used for endpoint agent tool calls. - The browser capability can rebuild agent activity as an auditable timeline. - The timeline includes the prompt, each tool call, approval decision, result and page-level evidence. - The timeline is correlated by browser tab and time. - Browser support runs through the same policy engine, inventory and audit pipeline as endpoint coverage. - Manifold says customers do not need a separate governance layer or a second set of policies. - Customers can deploy the extension through existing mobile device management tools. - The extension’s permission scope is limited to the AI surfaces it governs. - The extension does not read browsing history. - The only user data it reads is the signed-in account identity, which is used to enforce account policies. - Conversation content is not captured unless an organization explicitly enables prompt capture. - Monitoring an agent such as Claude in Chrome uses the same browser access the agent already relies on.

Between the lines: - The launch reflects a shift in where AI risk is concentrated, from endpoints alone to browsers that can serve as agent runtimes. - Manifold is positioning governance around observed behavior and evidence, not just prompts and tool access. - The company is using a low-permission deployment model to reduce friction for IT and security teams.

What’s next: - Manifold plans to add more browser platforms and AI surfaces throughout 2026. - The company is expanding from endpoint-only governance toward broader runtime coverage as agent usage spreads.

The bottom line: - Manifold is trying to make AI agents observable and controllable wherever they act, including inside the browser.