Lobe Hub

Your self-hosted AI workspace for chat, agents, and knowledge

Get it on umbrelOS

Version2.2.0

CategoryAI

Source codePublic

Developed byLobeHub 1 app

Submitted byal-lac 19 apps

Compatible withumbrelOS 0.5 or later

💬 Lobe Hub is a self-hosted AI workspace for conversations, agent teams, knowledge bases, and multimodal workflows.

🗣️ It supports speech, multimodal chat, files, knowledge retrieval, and a rich provider ecosystem including OpenAI-compatible APIs, Anthropic, Google, Groq, Ollama, and more.

🤖 This Umbrel package now follows the current upstream self-hosted deployment model and includes PostgreSQL, Redis, and S3-compatible object storage for uploads and knowledge base features.

🔑 To get started, create an account on first launch and then add one or more AI provider API keys in the Lobe Hub settings.

What's new

Version 2.2.0+1 day ago

This release introduces the Chief Agent Operator, an agent workflow that can self-review, iterate on its own output, dispatch sub-agent teams, and pause when human input is needed.

Key changes:

Added self-iteration, proposal actions, and auto-executed follow-ups for agent workflows
Added auto-formed sub-agent teams with monitor callbacks, breadcrumbed threads, and a thread switcher
Improved task scheduling, Claude Code task tools, and waiting-for-human status handling
Improved cloud Claude Code, GitHub token, and sandbox completion flows
Improved chat mode, the action bar, home recommendations, document tree updates, and MCP inspectors
Fixed bot identity gating issues for Slack, Discord, and device-tool activations

Full release notes are available at https://github.com/lobehub/lobehub/releases/tag/v2.2.0

Version 2.1.58+6 days ago

This hotfix release ships the latest canary improvements for agents, onboarding, integrations, desktop/CLI workflows, and reliability.

Key changes:

Improved agent tasks, self-review proposal flows, sub-agent dispatch, scheduling, and TodoList behavior
Polished home and onboarding flows, including daily brief links, input hints, and inline skill authorization
Fixed Slack, Discord, slash-command, and gateway client-tool integration issues
Improved desktop and CLI login/command detection, including Claude and Codex CLI handling on Windows
Added reliability safeguards for web crawling, image processing, database pools, and inactive OIDC access
Added the agent operations database table; self-hosted migrations run automatically on startup

Full release notes are available at https://github.com/lobehub/lobehub/releases/tag/v2.1.58

Version 2.1.57+last week

This release adds the generally available Agent Task System and improves agent workflows, inline documents, and bot integrations.

Key changes:

Added the Agent Task System for templates, task tracking, comments, scheduled runs, and dependency-ordered batches
Added inline document support for knowledge-base search and agent grounding
Improved heterogeneous agent execution and persistence across server replicas
Expanded bot platform support, including Messenger, Line, and Telegram fixes
Added a visual analysis tool and updated the default OSS DeepSeek model configuration

Full release notes are available at https://github.com/lobehub/lobehub/releases/tag/v2.1.57

Version 2.1.56+2 weeks ago

This release includes fixes from v2.1.55 and a database schema update in v2.1.56. The migration runs automatically on startup — no manual steps are required.

Key changes:

Fixed an issue where sending the first message in an empty Page Agent panel would clear the newly created topic
Fixed sending from the Home chat input not routing to the correct new topic when the route had no warm cache
Fixed stale topics being carried over when switching between agents or scoped pages
Added two new nullable columns (trigger, metadata) and an index to the briefs table — existing data is unaffected

Full release notes are available at https://github.com/lobehub/lobehub/releases

Version 2.1.54+3 weeks ago

This is a hotfix release addressing a navigation regression introduced in v2.1.53.

Key changes:

Fixed stale chat state when switching between agents — the previous topic no longer carries over
Fixed conversation context isolation so the correct agent/topic/thread state loads immediately on switch
Fixed sidebar thread list staying expanded while viewing a subtopic
Fixed the header showing the correct thread title for active subtopics

Full release notes are available at https://github.com/lobehub/lobehub/releases

Agent ZeroA personal, organic agentic framework that grows and learns with you

LibreChatEnhanced ChatGPT Clone

LocalAIDrop-in OpenAI replacement

OpenCodeThe open source AI coding agent

KokoroAn intelligent, high-quality TTS solution

OpenReaderA document reader with Text-to-Speech capabilities

Lobe Hub

Related