OpenClaw 2026.3.11-beta.1: The Beta You Install Because It Removes a Real Attack Path

This is a pre-release (beta). If you operate OpenClaw in production, treat it as a staged rollout.

Source: GitHub release — https://github.com/openclaw/openclaw/releases/tag/v2026.3.11-beta.1

1) Security fix: browser-origin WebSocket origin validation is now enforced (GHSA-5wcw-8jjv-m286)

If you run Gateway behind a reverse proxy (especially with any form of “trusted proxy” setup), this is the change that matters.

What the release says (paraphrased):

Gateway/WebSocket now enforces browser origin validation for all browser-originated connections, even when proxy headers are present.
This closes a cross-site WebSocket hijacking path in trusted-proxy mode that could grant operator.admin access.

Operational meaning:

If your Gateway is reachable from a browser and you rely on proxy headers, you should assume this is a high-priority patch.
After updating, verify that your legitimate dashboard/control UI still connects normally, and that unexpected origins cannot.

If you need an internal reference: GHSA-5wcw-8jjv-m286.

2) Breaking: isolated cron delivery becomes stricter (and a doctor fix is provided)

This beta tightens how isolated cron jobs are allowed to send notifications:

isolated cron jobs can no longer “notify” through ad hoc agent sends or fallback main-session summaries
migration support is added via openclaw doctor --fix for legacy cron storage and legacy delivery metadata

If you have cron-based reminders or reports:

Upgrade on staging first.
Run doctor and apply migration if it flags anything.
Trigger a sample job and confirm delivery still arrives in the expected channel.

3) Onboarding improvements that reduce support tickets

Ollama onboarding becomes first-class (Local or Cloud + Local)

The wizard adds:

Local mode and Cloud + Local mode
browser-based cloud sign-in
curated model suggestions
cloud-model handling that skips unnecessary local pulls

If you maintain a “local inference” setup for cost control, this reduces the number of ways onboarding can go sideways.

OpenCode: new “Go provider” + shared key handling

A new OpenCode Go provider lands, and the wizard/docs treat Zen and Go as one OpenCode setup while keeping runtime providers split.

Practical effect:

fewer mismatched keys across profiles
fewer “it worked yesterday but not in this profile” incidents

4) macOS chat UX: model picker + thinking level persistence

The macOS chat UI now includes:

a chat model picker
explicit thinking-level selections persisted across relaunch
hardened provider-aware model sync for the shared composer

If you do training/workflows with multiple providers, this is a small UX change that prevents big confusion.

5) Discord: auto-created threads can now set archive duration

If you use Discord auto-threading, a new channel config allows setting:

1 hour, 1 day, 3 days, or 1 week

instead of being stuck with the 1-hour default.

This matters if your community work happens over a few days and you keep losing threads.

6) Memory search: opt-in multimodal indexing (Gemini embedding model)

This beta adds optional image/audio indexing for memorySearch.extraPaths using gemini-embedding-2-preview, plus:

strict fallback gating
scope-based reindexing
configurable output dimensions and automatic reindex when dimensions change

Practical adoption guidance:

Turn it on only for specific paths (start small).
Expect a one-time reindex cost.
Validate search quality before rolling it into production workflows.

Recommended beta rollout checklist (ops-first)

Patch priority: if your Gateway is browser-reachable, prioritize this beta on staging due to the WebSocket origin fix.
Run openclaw doctor and apply --fix if cron migration is suggested.
Smoke test: dashboard connection, cron delivery, and your main channel plugin.
If you use Discord threads, set archive duration to match your community cadence.
If you enable multimodal memory indexing, scope it to a small dataset first.

FAQ: what should teams decide before they roll this beta wider

When does this beta deserve fast-tracking even for cautious operators?

It deserves fast-tracking when the security fix matches a real exposure pattern you already have, especially a browser-reachable Gateway behind reverse proxy or trusted-proxy assumptions. In that situation, the beta is not just an experiment, it is an early path to removing a concrete attack surface.

What is the wrong way to validate the cron changes in this beta?

The wrong way is to stop at a successful migration command or a clean release note read-through. The only validation that matters is whether a real scheduled job still lands in the exact destination your team depends on, without relying on undocumented fallback behavior.

Why should operators keep multimodal memory indexing scoped at first?

Because the main risk is not only build or provider failure, but operational sprawl. Once teams widen indexing scope too quickly, they can end up paying reindex cost and retrieval noise before they know whether the new memory surface is actually helping the workflows that matter.

OpenClaw News Editorial