Speed & Reliability
OpenClaw 2026.5.27 is a performance-focused release that streamlines Gateway operations, adds powerful new media and memory providers, and hardens delivery across all major messaging channels.
Performance: A Faster Gateway
We've optimized the core Gateway paths to reduce latency and CPU overhead, especially in environments with many active plugins or sessions.
- Metadata Caching: Session reads, plugin fingerprints, and auth env snapshots are now cached, reducing hot-path rediscovery.
- Streamlined Reply Paths: Visible replies no longer inherit hidden cleanup timeouts, leading to snappier response delivery.
- Lazy Discovery: Tool-search catalogs and plugin config are now handled more efficiently to avoid redundant startup cycles.
New Providers & Model Support
Expanding the ecosystem with more choices for embeddings, video generation, and model access.
- OpenAI-Compatible Embeddings: Core support for any OpenAI-style embedding provider, perfect for local or custom hosted endpoints.
- Pixverse Video Generation: Integrated Pixverse support with API region selection and full media generation workflow.
- DeepInfra Catalog: Improved onboarding with full model catalog loading and better pricing alignment.
- VLLM Thinking: Native support for VLLM thinking parameters in chat completion paths.
Channel Stability & Security
Hardening the connection between your agents and the platforms where they live.
- Durable Telegram Delivery: Outbound messages now use a more resilient path to survive transient connection drops.
- iMessage Refinement: Suppressed duplicate native execution approval prompts and improved attachment handling.
- Discord & Slack: Tighter requester checks for Discord and better cleanup handling for Slack final replies.
- Security Boundaries: Stronger prompt hygiene, normalized hostname handling, and restricted Node runtime environment overrides.
Codex & Developer Experience
Making the agent development and execution environment more robust.
- Resilient Codex Runs: Shared app-server clients and native hook relays now survive startup failures and restarts.
- Workspace Memory Routing: Improved memory routing through the Codex tool path for better context awareness.
- CLI Hardening: Rejected malformed numeric options and improved help target routing.
Technical Highlights
- Auth Profile Suffixes: Persisted model auth profile suffixes for more stable multi-account model routing.
- Diagnostic Events: Stabilized diagnostic event exports for easier troubleshooting.
- E2E Bounder: Bounded assertion waits across the entire QA suite to prevent CI wedging.
Upgrade Guide
Who should upgrade?
- Users seeking better Gateway responsiveness and lower CPU usage
- Developers working with Codex who need higher runtime reliability
- Teams using OpenAI-compatible local embeddings or Pixverse video generation
How to Upgrade
Update your installation via the terminal:
openclaw update
FAQ
What's new in Gateway performance?
We've implemented aggressive caching for session metadata, plugin fingerprints, and auth env snapshots. This significantly reduces hot-path rediscovery and makes the Gateway much more responsive.
Which new providers are supported?
This release adds core support for OpenAI-compatible embedding providers and the Pixverse video generation platform, including API region selection.
What improvements were made to Telegram delivery?
Telegram sendMessage actions now use a more durable outbound delivery mechanism, ensuring messages are reliably sent even during transient network issues.
How does the Codex reliability update help?
Codex app-server runs are now more stable with better shared client management and resilient hook relay generations that survive restarts.
Need help from people who already use this stuff?
Join the OpenClaw Community
Get help with your setup, share your agent workflows, and stay up to date with the latest releases in our profit-lab community.