OpenClaw v2026.10.2 Patch Release Fixes Upgrade Issues and Reply Errors

·2026.10.13 02:11

Key point

The update includes 80 PRs and 12 direct commits from 107 contributors, addressing critical bugs in upgrades, message delivery, and long conversation stability.

Details

OpenClaw v2026.10.2 has been released as a patch update, focusing on stability improvements for upgrades, message handling, and long-running conversations. The release incorporates 80 PRs and 12 direct commits from 107 contributors.

Upgrade and System Stability

The patch resolves several issues related to upgrading from version 2026.10.1, including cleanup errors after package replacement and a final plugin check that could block Windows upgrades. It also improves Doctor repairs by bypassing false failures, fixes stale Gateway locks after container replacement, and addresses false maintenance warnings and Linux database checks.

Message Delivery and Conversation Handling

Fixes target missing or duplicate replies in affected chats, including dashboard replies after daily or idle resets, missing final reports from long Claude CLI tasks, and replies lost after helper results. Long conversations are now more resilient, with configured fallbacks handling compaction while primary credentials cool down. Corrections sent during an answer remain in the stored context when using a durable context engine, and Telegram previews stay visible until their replacement is accepted.

Performance and Security Fixes

Credential masking now handles long tool output more reliably, fixing issues with recognized credentials split across output and preventing freezes or stack overflows on large or repetitive text. Additional fixes address background tasks incorrectly cancelled as stalled, plugin tools timing out while waiting for approval, and chat freezes caused by loading large plugins twice.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.