OpenClaw 2026.5.5 Release Notes: xAI Grok Fixes, Telegram Progress Drafts, and LINE Validation

OpenClaw v2026.5.5 was released on May 6, 2026. The official release page shows a focused maintenance update centered on channel routing, xAI compatibility, approval delivery, and Control UI session presentation.

This release sits between the broader OpenClaw 2026.5.4 feature work and the narrower OpenClaw 2026.5.6 Codex OAuth recovery patch.

xAI Grok Compatibility Fixes

Two of the most important fixes in 2026.5.5 are for xAI-backed models.

The official release notes say OpenClaw stopped sending OpenAI-style reasoning effort controls to native Grok Responses models, specifically so xai/grok-4.3 no longer fails live Docker or Gateway runs with Invalid reasoning effort. The release also says the bundled xAI thinking profile is now clamped to off so unsupported reasoning levels are not sent to native Grok Responses models.

The official xAI provider docs confirm that OpenClaw uses the xAI Responses API as the bundled xAI transport and lists grok-4.3 in the built-in model catalog. For users actively running Grok through OpenClaw, that makes this release operationally relevant.

Telegram, Codex, and Progress Draft Behavior

The May 6 release notes also include a user-facing fix for progress visibility. OpenClaw now keeps message-tool-only progress drafts visible in Telegram and renders native Codex tool progress once per tool instead of duplicating draft lines.

That aligns with the official progress drafts documentation, which describes one visible work-in-progress message that updates while the agent reads, plans, waits, or uses tools. The 2026.5.5 change is not a new feature. It is a correction so Telegram and Codex progress behavior better matches that design.

Channel and Approval Reliability

Several other fixes in this release tighten behavior across messaging platforms:

Feishu now hydrates missing native topic starter thread IDs before session routing so first turns and follow-ups stay in the same topic session.
LINE now rejects dmPolicy: "open" configurations that do not include wildcard allowFrom, instead of acknowledging webhook DMs and then blocking them silently.
Matrix approval delivery is retried up to three times with a short backoff so transient send failures do not leave approvals stranded.
Discord heartbeat ACK timeout measurement now starts from the actual heartbeat send, which the release notes say prevents false reconnect loops while the channel is still awaiting readiness.
Plain text control commands such as /steer are now routed through normal authorization and mention gating in Discord instead of being dropped before an agent session can see them.

Control UI Session Changes

One additional item in the release notes affects the session experience in the Control UI. OpenClaw now presents the compaction count as a compact disclosure and expands session-level details using checkpoint history cards across responsive table layouts.

That is a presentation change, but it is still useful for operators inspecting session history and compaction behavior after long-running agent turns.

Why This Release Matters

OpenClaw 2026.5.5 is the kind of release that stabilizes edge cases users actually hit:

xAI Grok runs failing because of unsupported reasoning fields
progress drafts becoming noisy or duplicated
webhook policies that appear valid but silently block delivery
chat sessions losing continuity in threaded channels

If your deployment spans Telegram, Discord, LINE, Matrix, Feishu, or xAI models, v2026.5.5 is a meaningful compatibility update.