OpenClaw🦞
@openclaw
OpenClaw 2026.4.26 🦞 🎙️ Google Live Talk 🦙 Better Ollama/local models 🧳 Bring over Claude + Hermes setups 🔐 One-command Matrix E2EE Big release. Local models eat well. https://t.co/90QZLfSUhC
65retweets652likes
View on XOpenClaw v2026.4.26 Ingests Claude Code and Hermes Agent Setups
OpenClawOpenClaw v2026.4.26 introduces a migration tool to import configurations and memory from proprietary agents like Claude Code directly into its self-hosted environment. The update also adds Google Live Talk for real-time voice and optimizes local model retrieval for Ollama embedding backends.
OpenClaw, a self-hosted AI assistant that connects models to messaging platforms, released version 2026.4.26 with a migration tool. This tool imports configurations and MCP servers from Claude Code, Claude Desktop, and Hermes. The release also adds Google Live Talk for real-time voice.
This update shifts OpenClaw into an orchestrator capable of absorbing existing agentic workflows. By bridging proprietary, cloud-tethered tools to private infrastructure, it addresses data sovereignty without sacrificing ecosystem progress. It also hardens Active Memory plugins to ensure local performance remains competitive with cloud-based alternatives.
You can now run openclaw migrate to move your agent "soul" into a self-hosted instance. The update includes a one-command setup for Matrix end-to-end encryption and adds Cerebras as a bundled provider. For local models, the system now applies optimized retrieval prefixes for nomic-embed-text and qwen3-embedding.
Still wondering? A few quick answers below.
Users can migrate by running the openclaw migrate command in the terminal. This tool supports importing configurations, memory hints, model providers, and MCP servers from Claude Code, Claude Desktop, and Hermes. It includes a plan and dry-run mode to ensure a safe transition from these proprietary agent environments to self-hosted infrastructure.
Google Live Talk is a new real-time voice transport layer integrated into the OpenClaw Control UI. It allows for low-latency browser-based voice sessions using ephemeral tokens and a gateway relay. This feature enables users to interact with their self-hosted agents through voice while maintaining a stable connection via the WebSocket transport protocol.
OpenClaw v2026.4.26 introduces model-specific retrieval prefixes for local embedding models like nomic-embed-text and qwen3-embedding. These prefixes improve memory search accuracy by tailoring queries to the specific requirements of each model. Additionally, the update streamlines the onboarding process by deduplicating suggested models against existing local installs to prevent redundant downloads.
You can enable Matrix end-to-end encryption by running the openclaw matrix encryption setup command. This single-command flow handles the entire process, including bootstrapping recovery keys and printing the final verification status. This ensures that your communications across the Matrix messaging platform remain private and secure within your self-hosted AI assistant environment.
The latest update adds full group chat support for QQBot, including history tracking and per-group configurations. It also registers the Tencent Yuanbao external channel plugin for bot direct messages and group chats. These integrations allow users to deploy their AI agents into the Tencent social ecosystem with unified media sending and streaming message capabilities.
Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →
