An open-source, modern-design AI chat framework. Supports Multi AI Providers (OpenAI / Claude 4 / Gemini / Ollama / DeepSeek / Qwen), Knowledge Base (file upload / knowledge management / RAG), Multi-Modals (Plugins/Artifacts) and Thinking.
What's new in 2.2.20
This release was automatically published from PR #21078. Changes See PR description: https://github.com/lobehub/lobehub/pull/21078 Commit Message 🚀 LobeHub Release (20261018) Release Date: October 18, 2026 Since v2.2.19: 419 merged PRs · 14 contributors Agents now reach further across your devices, and the app opens faster. Topics, documents and device data load from local copies first, and server runs can use models that only exist on your device.
✨ Highlights
Goals with batch rollout — A goal stays a draft until you start it, and a Goal can roll out one repeated task in batches. Each batch stops at a decision gate for the owner to review. (#20840, #20704, #20917, #20833)
Widgets & dashboards — Home and project dashboards, an agent widget tool, chat previews, and new lh widget / lh dashboard commands. (#20421, #20771)
LLM relay to your device — Server-side runs, preset tasks, model lists and Ollama downloads can now use models that are only available in your browser tab or desktop app. (#20453, #20454, #20455, #20456, #20457)
Remote terminals & device maintenance — Open interactive shells on connected devices from the web, and update or restart CLI daemons from device settings. (#20600, #20527, #20528)
Local-first loading — Topics, agent documents, devices, git status and home recents load from local replicas first, so pages show content before the network responds. (#20449, #20869, #20838, #20841, #20851)
Agent document folders — Folders, batch moves, documents as parents, and sorting by name, created or modified time. (#20761, #20926)
Memory from your materials — Memory can extract rules from pasted text, files, documents and conversations. (#20266)
Fork from a reply — Start a new topic from any assistant message. (#20610)
🏗️ Core Agent & Architecture - Sub-agents run through the parent's executor, and afterToolCall controls can gate tool results. (#20446, #20297) - Skills selected in the composer or CLI are injected on server runs. Adding an agent installs the Market skills it references. (#20961, #21059) - Agent-to-agent messages now show which agent sent them. (#20548) - Fewer repeated device and Market lookups on each send, and project sends appear right away. (#20465, #20897, #20628) 📱 Platforms / Integrations - Heterogeneous agents: DeepSeek Harness adapter. Codex app-server can run on connected devices (behind Labs). Codex user messages can be edited and resent. Codex and Pi can use any tool-capable LobeHub model. …
Unleash Claude’s raw power directly in your terminal. Search million-line codebases instantly. Turn hours-long workflows into a single command. Your tools. Your workflow. Your codebase, evolving at th
Codex CLI is an open‑source local coding agent that runs in your terminal, letting you write, edit, and understand code without leaving the command line.