OpenClaw adds a new Active Memory plugin, automatically retrieving memories during conversations without manual triggers

ME News Report, April 13 (UTC+8), according to 1M AI News monitoring, the open-source AI agent tool OpenClaw released version v2026.4.12, with the main highlight being the addition of the Active Memory plugin.
Active Memory is an optional plugin that automatically runs a dedicated memory sub-agent before each reply, proactively retrieving user preferences, historical context, and past details related to the current conversation, without requiring the user to manually input “remember this” or “search memory.”
The plugin supports three context modes (single message, recent messages, full conversation) and allows real-time viewing of the memory retrieval process via the /verbose command.
Other major new features include:

1. macOS Talk Mode adds an experimental local MLX speech engine, enabling speech synthesis to run locally without calling cloud APIs, with automatic fallback to system speech when unavailable.
2. Built-in Codex model provider, allowing \codex/gpt-*\ models to be used directly through native Codex authentication and thread management.
3. Built-in LM Studio provider, supporting locally hosted and self-managed OpenAI-compatible models, including automatic model discovery and embedding vector search.
4. Added openclaw exec-policy command for local management of execution approval policies.
5. Plugin loading optimization, loading only necessary components declared in the manifest at startup to avoid loading unrelated plugins during runtime.
   Security fixes include three vulnerabilities: preventing bypass of authorization checks via empty approver lists, removing busybox/toybox from the list of secure binaries to prevent interpreter escape, and enhancing shell wrapper detection to prevent environment variable injection attacks.
   (Source: BlockBeats)