Summary
OpenClaw versions 2026.1.5 prior to 2026.2.12 expose the /agent/act browser-control HTTP route without enforcing mandatory authentication.
This allows local network attackers or local processes to invoke privileged browser-context actions and access sensitive in-session data by sending requests to unauthenticated endpoints.
Affected context
The vulnerability appears in OpenClaw deployments that expose the /agent/act browser-control HTTP route and rely on it for privileged actions such as driving or inspecting the browser.
It becomes relevant when:
- The service is reachable from the local network or from other local processes on the same host
- No authentication, session binding, or origin checks are enforced on the control endpoint
- Browser sessions with sensitive, logged-in user data are active and controlled through this route
Exploit preconditions
- Attack surface: HTTP interface at the
/agent/actbrowser-control route - Required access: Network access to the OpenClaw service from the local network, or ability for a local process to reach the bound interface/port
- Trigger point: Sending crafted HTTP requests to the unauthenticated
/agent/actendpoint to invoke browser actions - Why exploitation is possible: The route performs privileged operations without any authentication, authorization, or session validation, so any reachable caller can trigger sensitive actions as if they were the legitimate controlling user
Why this happens
The core issue maps to CWE-306: Missing Authentication for Critical Function.
A critical browser-control function (/agent/act) that can operate on live browser sessions and their data is exposed without requiring credentials or verifying the caller's identity.
This design implicitly trusts any reachable client and does not bind the control channel to a specific user, session, or origin, allowing unauthorized callers to exercise powerful capabilities.
Potential impact
- Unauthorized execution of browser-context actions (navigation, clicks, form submissions, script execution)
- Access to sensitive in-session data, including authenticated application pages and potentially cookies or tokens visible in the browser context
- Data exfiltration from web applications the browser is logged into
- Manipulation of user workflows, transactions, or approvals conducted through the controlled browser
- Use of the browser instance as a pivot point to interact with internal web applications the attacker could not otherwise reach
General mitigation guidance
Effective mitigation centers on robust authentication and authorization for all privileged endpoints:
- Require strong authentication (e.g., API keys, tokens, or mutual TLS) for all browser-control routes, including
/agent/act - Bind control endpoints to a specific authenticated user or session context and validate that association on every request
- Restrict network exposure of control interfaces to trusted hosts or loopback-only where feasible
- Enforce least privilege in the API design so that only the minimum set of actions is available to any authenticated caller
- Implement strict input validation and action whitelisting on control commands to reduce inadvertent misuse
How teams detect issues like this
Organizations can detect similar issues using a combination of design review and automated checks:
- Static analysis / code scanning: Search for HTTP routes that trigger high-privilege operations without accompanying authentication or authorization checks
- Configuration and architecture review: Identify services that expose control interfaces on non-local interfaces or to untrusted networks
- Access control policy audits: Verify that all routes capable of session interaction, browser control, or data extraction have explicit access-control middleware or filters applied
- Runtime monitoring: Log and inspect calls to browser-control endpoints for unexpected sources, patterns, or volumes, and flag unauthenticated access attempts

