OpenClaw v2026.4.12, released on April 13, 2026, is a broad quality release focused on plugin loading, memory and dreaming reliability, new local-model options, and a much smoother Feishu setup path.
Active Memory Plugin
The new optional Active Memory plugin gives OpenClaw a dedicated memory sub-agent that runs right before the main reply. Ongoing chats can automatically pull in relevant preferences, context, and past details without users having to manually say "remember this" or "search memory." Active Memory supports configurable message, recent, and full context modes, live /verbose inspection, advanced prompt and thinking overrides, and opt-in transcript persistence for debugging.
Bundled Codex Provider
OpenClaw now ships a first-class Codex provider with a plugin-owned app-server harness. Models referred to as codex/gpt-* use Codex-managed auth, native threads, model discovery, and compaction, while openai/gpt-* stays on the normal OpenAI path. Operators who want to run Codex alongside standard OpenAI runs no longer have to choose between them.
LM Studio Provider
A new bundled LM Studio provider adds onboarding, runtime model discovery, stream preload support, and memory-search embeddings for local and self-hosted OpenAI-compatible models. Combined with Ollama, OpenClaw now has two bundled paths for purely local operation.
New exec-policy CLI
The new openclaw exec-policy command, with show, preset, and set subcommands, synchronizes requested tools.exec.* config with the local exec approvals file. Follow-up hardening rejects node-host elevation, adds rollback safety, and detects sync conflicts so approval state and config cannot silently drift.
Plugin Loading Narrowed
Plugin activation now narrows to manifest-declared needs and centralizes manifest-owner policy, so startup, command discovery, and runtime activation avoid loading unrelated plugin runtime. A Convex-backed pooled Telegram credential leasing system and openclaw qa credentials admin commands land for QA lab workflows.
Security & Fixes
The release hardens shell wrapper detection, blocks env-argv assignment injection, strips BusyBox/Toybox from interpreter-like safe bins, and prevents empty approver lists from granting approval authorization. Dreaming fixes keep managed heartbeats from consuming twice and stop dreaming narratives from re-ingesting themselves. Deploy the release via OpenClaw Launch.