By devasher · Edited by Nominiclaw
OpenClaw v2026.6.8 introduces richer Telegram and WhatsApp delivery, expanded provider support including GLM-5.2 and Kimi K2.7 Code, and significant stability improvements for agent recovery and memory indexing.
OpenClaw v2026.6.8 delivers enhanced rich-media delivery for Telegram and WhatsApp, expanded LLM provider support, and hardened recovery mechanisms for agents, gateways, and memory indexing.
Telegram and WhatsApp delivery are now more robust and feature-rich. Telegram supports structured rich text, including tables, lists, and expandable blockquotes, while preserving intentional line breaks. WhatsApp now honors configured ACP bindings. Additionally, Slack now emits message_sent hooks on outbound delivery, mirroring Telegram's behavior.
Model handling has been expanded with the addition of GLM-5.2 and Kimi K2.7 Code (which features always-on reasoning). Claude Haiku 4.5 catalog entries have been added, and provider-prefix normalization has been tightened for OpenRouter and Google Vertex paths to prevent invalid API requests.
Recovery mechanisms have been sharpened across several critical paths, including account-scoped DM sends, generated media completions, and heartbeat deduplication. The Gateway now marks active main sessions before restart shutdown aborts to ensure better restart recovery. Additionally, the /btw command is now supported in CLI-backed sessions, allowing for ephemeral side-questions without polluting main chat history.
Memory indexing is now more resilient; oversized OpenAI embedding batches are split before hitting 431 errors, and full memory reindexes now preserve retry state and embedding cache during rollbacks. SQLite now avoids WAL on NFS state volumes to prevent corruption during container revision replacements.
WebChat now preserves backscroll during streaming, and the sidebar session picker remains interactive above the desktop workbench. The Workspace Files panel now starts collapsed by default to maximize conversation width. On iOS, stale foreground gateways are now forced to reconnect more reliably.
nominiclaw/nodes documentation now includes a comprehensive openclaw.json example. openclaw doctor now provides actionable warnings for untrusted external Discord plugins and identifies Slack channel-map keys that cannot be routed./usage full footer renderer is available in core, allowing for highly customized usage bars without requiring a renderer plugin./v1/responses are now forwarded to agents, achieving parity with /v1/chat/completions.message tool now ensures that contributed schema properties remain optional, preventing validation failures across different channels.qa-evidence.json artifact and includes a simplified scorecard mapping shape.This release introduces rich-text support for Telegram, ACP binding support for WhatsApp, new models like GLM-5.2 and Kimi K2.7 Code, and improved recovery for agents and memory indexing.
No breaking changes are reported in the source material.
No specific migration steps or breaking changes are mentioned; users can upgrade via standard update paths.
Deploy your own OpenClaw bot in under 120 seconds — no server, no code.
Deploy OpenClaw →