OpenClaw Weekly Changelog

ABC delivered

This week Vincent had the rains.

Stability & security

• GPT5.4 Behaviour Tuning
• Active Memory Plugin
• Multimodal is becoming real product surface area
• Provider routing and recovery are getting more serious
• Microsoft teams got some love
• Security hardening keeps showing up everywhere

Strict mode and native Codex as harness

agents.defaults.embeddedPi.executionContract = "strict-agentic"

• This tells GPT-5.x to keep working: read more code, call tools, make changes, or return a real blocker instead of stopping at “here's the plan.”

plugins.entries.codex.enabled = true
agents.defaults.model = "codex/gpt-5.4"
agents.defaults.embeddedHarness = {
  runtime: "codex",
  fallback: "none"
}

• Codex owns threads, resume, compaction, app-server execution.

Memory is still the strategic bet

• Active memory gives OpenClaw a dedicated memory sub-agent right before the main reply, so ongoing chats can automatically pull in relevant preferences, context, and past details without making users remember to manually
• Dreaming and recall systems got refined
• More work on session recovery and compaction resilience
• The product is clearly being shaped around long-horizon usefulness

Multimodal is no longer a side quest

• Video and music generation keep expanding the product surface
• add Seedance 2.0 model
• Gemini TTS support landed in the Google plugin
• OpenClaw keeps moving from agent shell to broader workbench

Inference and fallback got more serious

• Better provider routing matters more than model-picking theater
• Fallback behavior is becoming a real product feature
• Session restore and recovery work reduce dead-end failures

Security keeps showing up for a reason

• SSRF tightening
• auth and approval hardening
• replay and transport safeguards
• command and path safety fixes

What contributors were actually feeling

• PR review speed is part of developer experience now
• Docs gaps still create avoidable friction
• Plugin, provider, and channel edge cases keep burning time
• Contributors are still trying to ship, which is the healthy signal underneath the complaints

A few concrete releases worth saying out loud

• Anthropic defaults moved to Claude Opus 4.7
• Gemini TTS support landed in the bundled Google plugin
• Memory LanceDB cloud storage support landed
• macOS app nodes got screen.snapshot

Do not read this like a grocery list

• Lead with the pattern
• Spend more time on why it matters than naming every patch
• Keep the room focused on direction, not changelog fatigue
• Cleaner line: broader product, deeper memory, safer runtime.

OpenClaw is getting more useful in the ways serious software usually gets more useful

• better memory
• better routing
• better recovery
• broader media capability
• more security work in the seams