OpenClaw 2026.4.12 adds automatic memory recall and fixes a long list of reliability issues

ByIvan Jenic

OpenClaw 2026.4.12 ships Active Memory, a new sub-agent that automatically pulls in relevant context before each reply, alongside a wide batch of stability fixes.

openclaw logo

OpenClaw 2026.4.12 is out. The release is mostly stability work, but it ships one meaningful new feature: Active Memory, an optional plugin that gives OpenClaw a dedicated memory sub-agent.

The sub-agent runs before the main reply and automatically pulls in relevant preferences, context, and past details from memory without requiring users to manually trigger recall. Previously, getting OpenClaw to surface relevant memory meant explicitly asking it to search or remember. Active Memory makes that happen on its own.

It ships with configurable modes for message, recent, and full context, along with a live /verbose inspection command and opt-in transcript persistence for debugging. Documentation is at docs.openclaw.ai/concepts/active-memory.

Other additions in this release:

  • Local speech on macOS: An experimental MLX speech provider for Talk Mode, with explicit provider selection, local utterance playback, and a system-voice fallback
  • Gateway command discovery: A new commands.list RPC lets remote gateway clients discover available runtime, text, skill, and plugin commands
  • exec-policy CLI: A new openclaw exec-policy command for syncing requested tool execution config with the local approvals file

The fix list is long. Among the things addressed:

  • Telegram deadlock: Approval button callbacks now route on a separate lane so plugin approvals resolve immediately instead of blocking behind an active agent turn
  • Dreaming timezone: Diary timestamps now use the host’s local timezone when dreaming.timezone is unset, with the abbreviation included in DREAMS.md
  • WhatsApp media: Outbound media sends no longer silently drop attachments when mediaUrl is empty but a resolved media list exists
  • Audio transcription: The real provider failure now surfaces when transcription attempts fail, instead of collapsing into a generic skip
  • Security patches: Three fixes covering interpreter safe-bin hardening, empty approver list auth bypass, and shell-wrapper injection blocking
  • Agents/queueing: Follow-up messages that arrived mid-run were being silently dropped; they now carry into the next prompt correctly
  • WebSocket keepalive: Tick broadcasts no longer get marked as droppable, which was causing slow clients to self-disconnect during long-running tasks

Source: OpenClaw/GitHub

Source: OpenClaw/GitHub

RunPod
RunPod

If you need on-demand GPUs for training, fine-tuning, inference, or running open-source models, give RunPod a try.

  • Available hardware: H100, H200, A100, L40S, RTX 4090, RTX 5090, and 30+ more
  • Cost: significantly cheaper than AWS or GCP, billed per second, no contracts
  • Setup: spins up in under a minute, 30+ regions worldwide
Try RunPod →
Affiliate disclosure: We may earn a commission if you sign up via our link, at no extra cost to you.
Efficienist Newsletter

Get the core business tech news delivered straight to your inbox. We track AI, automation, SaaS, and cybersecurity so you don't have to.

Just read what you want, and be done with it.

Read Next