OpenClaw 4.1 in Focus: SearXNG Web Search, Voice Wake, and the Cron Tools Allowlist

Release

OpenClaw 2026.4.1 Release: SearXNG, Voice Wake, and Bedrock Guardrails Land

The 4.1 release adds SearXNG as a bundled self-hosted web search provider, which is a big deal for operators who want search without routing queries through a third-party API. Voice Wake for Talk Mode and Amazon Bedrock Guardrails support round out a release that covers infrastructure, UX, and enterprise compliance in one shot.

Tasks Get a Chat-Native Board with /tasks and Agent Failover Gets Rate-Limit Caps

The new /tasks command surfaces a live background task board inside any chat session, which closes the gap between what the agent is doing and what you can actually see. The rate-limit failover cap added alongside it means agents on heavy workloads stop burning through auth profiles before trying a different provider.

Tools

Cron Gets Per-Job Tool Allowlists via openclaw cron --tools

You can now scope exactly which tools a scheduled cron job is allowed to call, which is the kind of granular control that makes automated agent jobs safe to run without babysitting. This pairs directly with the exec hardening from 3.31 and gives operators a real lever to limit blast radius on background work.

Appcast Update Confirms 3.31 Stable: Browser Attach Fix and Feishu Media Sends Patched

The stable 3.31 appcast push includes fixes for Chrome MCP attach timeouts on macOS and Feishu file attachment sends that were silently failing schema validation. If you skipped the beta and are updating now, these fixes are in and the six breaking changes from 3.31 are live.

Analysis

OpenClaw 3.28: Plugin Approval Hooks and ACP Channel Binds Explained

The async requireApproval hook in beforeToolCall lets plugins gate any tool execution behind a human confirmation, routed through whatever channel is active including Telegram buttons and Discord interactions. ACP channel binds for Discord and iMessage mean you can turn an existing chat into a Codex workspace without spawning a child thread.

Exec Obfuscation Detection in 3.31 Will Block Your Own Agent Commands

The detectCommandObfuscation() function added in 3.31 flags legitimate commands as suspicious even after setting ask: off, which caught a lot of operators off guard in production. If you are on any version from 3.31 onward and see unexpected exec blocks, this is the culprit and you need an explicit policy in exec-approvals.json.