Island Enterprise Browser for Agents

Island Enterprise Browser for Agents is Island's enterprise browser built for AI agents, announced October 29, 2026. It lets agents research the web and operate SaaS/internal apps under the same policy, identity, credential isolation and audit model Island already applies to human workers—on device and in the cloud, without relying on an outside hosted browser.

Official: island.io press · Company: island.io

Key points

1. Named product line: not just "Island browser supports agents"—press titles it Enterprise Browser for Agents.
2. Governed browser tools: per-agent/per-app allowlists down to DOM operations and read vs write; MCP layer is policy-governed too.
3. Identity + attribution: agents inherit enterprise identity with separate agent policy; every event records human vs agent for SOC 2 / ISO / HIPAA / PCI / EU AI Act-style workflows; optional full session recording.
4. Safety primitives: read-only mode on systems of record, data boundaries, vault-injected credentials that never enter the model context, ZTNA for internal apps without VPN.
5. Efficiency: annotation-based page APIs (Island owns the engine) to cut screenshot/token waste; learns repeated navigation paths.

Why it matters

Distinctive category name for "enterprise browser for agents" shipping from a known SEB vendor this window.

Who it's for

Security and platform teams already on Island (or evaluating secure enterprise browsers) that need agents to browse CRM, ERP and internal apps without ad-hoc consumer browser stacks.

Contact Island sales / [email protected] for deployment on endpoints and cloud agents.