Your super agent for work: local-first, learn your working context in mins and never forget it.
HolaOS is a local-first desktop agent runtime built with Electron and TypeScript that connects to over 100 workplace tools, including Linear, GitHub, Slack, Jira, HubSpot, and Gmail, via one-click OAuth to automatically fetch context and build a persistent working memory without requiring the user to re-explain background on each session. It integrates with Claude and other large language models through an orchestrator-plus-subagent architecture, where a user-facing agent manager coordinates hidden subagents running tasks in parallel and surfaces consolidated results through a single desktop UI. Memory is stored locally as Markdown files, embedded using SQLite vec, and retrieved through RAG, an approach inspired by Andrej Karpathy's LLM wiki workflow. A session compaction system reserves roughly 70 percent of the model context window for active reasoning by folding older history into structured checkpoints that preserve goals, decisions, and progress. The project targets knowledge workers who need persistent, auditable, locally controlled agent memory across long-running projects.
- ✓License: NOASSERTION
- ✓Actively maintained (<30d)
- ✓Healthy fork ratio
- ✓Clear description
- ✓Topics declared
- ✓Documented (README)
git clone https://github.com/holaboss-ai/holaOS && cp holaOS/*.md ~/.claude/agents/11 items in this repository
Build a new holaOS app using @holaboss/app-builder-sdk (5 backend primitives + optional shadcn dashboard UI). The canonical path for vibe-coded apps — integration modules AND dashboard apps both live here.
Use when working in the embedded browser and you want the cheapest reliable interaction loop.
Use when validating or debugging a workflow in the embedded browser and you need a reproducible, evidence-first loop.
Build the visual layer of a holaOS dashboard app — TanStack Start + @holaboss/ui + workspace tokens. Use when an app has SDK primitives wired (via app-builder-sdk) AND needs a `src/client/` UI surface. NOT for marketing pages, NOT for snapshot HTML reports.
Provision a production-ready teammate only after its stable responsibilities, prerequisites, and reusable operating guidance are understood.
Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, artifacts, posters, or applications. Generates creative, polished code that avoids generic AI aesthetics.
This skill is for interface design — dashboards, admin panels, apps, tools, and interactive products. NOT for marketing design (landing pages, marketing sites, campaigns).
Guide for creating effective skills. Use when creating or updating a skill.
Remove AI writing patterns from prose. Use when drafting, editing, or reviewing text to eliminate predictable AI tells.
Subagents overview
<p align="center">
<img src="docs/images/banner.png" alt="Holaboss logo" />
</p>
<p align="center"><strong>Your super agent for work: local- first, learn your working context in mins and never forget it</strong></p>
<p align="center">
<a href="https://github.com/holaboss-ai/holaOS/actions/workflows/ci.yml"><img src="https://github.com/holaboss-ai/holaOS/actions/workflows/ci.yml/badge.svg" alt="CI" /></a>
<img src="https://img.shields.io/badge/node-24.14.1-43853d" alt="Node 24.14.1" />
<img src="https://img.shields.io/badge/platform-macOS%20supported,%20Windows%20%26%20Linux%20in%20progress-f28c28" alt="macOS supported, Windows and Linux in progress" />
<img src="https://img.shields.io/badge/desktop-Electron-47848f" alt="Electron desktop" />
<img src="https://img.shields.io/badge/runtime-TypeScript-3178c6" alt="TypeScript runtime" />
<img src="https://img.shields.io/badge/license-Modified%20Apache%202.0-0f7ae5" alt="Modified Apache 2.0 license" />
</p>
<p align="center">
<a href="https://x.com/Holabossai"><img src="https://img.shields.io/badge/Follow%20on%20X-000000?style=for-the-badge&logo=x&logoColor=white" alt="Follow on X" /></a>
<a href="https://discord.com/invite/NSeHUCBj6"><img src="https://img.shields.io/badge/Join%20our%20Discord-5865F2?style=for-the-badge&logo=discord&logoColor=white" alt="Join our Discord" /></a>
</p>
<p align="center"><strong>⭐ Help us reach more developers and grow the holaOS community. Star this repo!</strong></p>
<p align="center">
<a href="https://www.holaos.ai/?utm_source=github&utm_medium=oss&utm_campaign=hola_boss_oss&utm_content=readme_nav_website">Website</a> ·
<a href="https://www.holaos.ai/docs/getting-started?utm_source=github&utm_medium=oss&utm_campaign=hola_boss_oss&utm_content=readme_nav_docs">Docs</a> ·
<a href="https://www.holaos.ai/signin?utm_source=github&utm_medium=oss&utm_campaign=hola_boss_oss&utm_content=readme_nav_signin">Sign in</a> ·
<a href="#quick-start">Quick Start</a>
</p>
# HolaOS (What is HolaOS):
<p align="center">
<img src="docs/images/desktop-workspace.png" alt="holaOS desktop workspace screenshot" width="1280" />
</p>
<!-- <p align="center">
<em>Inside the workspace: apps, files, a completely customizable dashboard, and agent chat live side by side while a single user-facing agent manager coordinates sub agents in the background.</em>
</p> -->
HolaOS is your super agent for work: local-first, learn your working context in mins and never forget it, below are the key features:
- **With 100+ Integration and Auto-fetch**, Agent know your state of work in mins
With one-click OAuth, your Agent connects to the tools, browser profile, local files you already use every day — Linear, GitHub, Slack, Jira, HubSpot, Gmail, and 100+ other workplace integrations.
No more copying updates, pasting links, or re-explaining background. Agent automatically fetches (only when you allowed) relevant signals from your tools in mins and turns scattered app data into working memory. Agent brings the right context back when you need it.
- **Memory Management system** that let your Agent never forget your working memory
Your Agent never starts work from zero — it remembers your working context, where the work stands, and brings the right memory back when you need to continue. The agent builds a local-first, compressed knowledge base from workspace files, browser state, integrations, and work activity, turning raw data from tools like Gmail, Slack, Notion, GitHub, and Jira into durable memory. Inspired by Karpathy’s LLM wiki workflow, this memory is stored locally as Markdown, embedded with SQLite vec, and retrieved through RAG, preserving key facts and context while making recall faster, more controlled, and easier for both you and the agent to use. You can open, browse and edit.
- **Session Context compression** that keep your working context fresh and token efficiency
The Agent remembers the past without sacrificing its ability to think about the current task. By our Safe Session Compaction: long-running agents are only useful if they stay coherent after days of work, not just one oversized prompt. holaOS keeps roughly 70% of the model window reserved for fresh reasoning, preserves the active working set verbatim, and folds older history into structured checkpoints that retain goals, constraints, progress, decisions, next steps, critical context, and file activity. It cuts history at sensible boundaries when possible, repairs split turns with local prefix summaries when necessary, and
performs compaction on a snapshot that only merges back if the live session still matches. The result is durable continuity without the usual slow drift into bloated, expensive, fragile sessions.
- **Simple UI & User friendly**
Use agents through a desktop app, not a terminal. You can manage tasks, inspect files, view outputs, and interact with your Agent in one simple UI — no CLI required. A local file system keeps work artifacts out of long conversations, so plans, drafts, notes, generated files, configs, and outputs stay in a visible workspace that both you and the Agent can access and edit. With one account, you can use leading SOTA models without managing separate providers, API keys, or setup. For complex tasks, hidden subagents can work in parallel while the orchestrate agent brings back progress and deliverables as one easy-to-review result. Integrate your Browser profile with one-click to support Agent handle real web-based work across websites, dashboards, and apps beyond standard APIs or integrations.
- **Memory never out of you control**
Your Agent learns your work, but your working memory stays in your computer. Built local-first, your data is not locked inside someone else’s cloud.
You can see what the Agent remembers, edit, remove what should not be kept, and control what gets fetched or recalled. Memory is always visible, adjustable, and owned by you.
# Working Agent Memory mechanism
The agent creates a local-first knowledge base from the files, browser state, integrations, and work activity connected to your workspace. Information from tools like Gmail, Slack, Notion, GitHub, and Jira (one - click OAuth) is summarized and compressed before becoming memory, so the agent does not need to search through every raw email, message, document, or ticket from scratch.
Memory is stored locally as Markdown files and embedded with SQLite vec on your machine. This gives the agent a durable structure it can retrieve from through RAG, while keeping the underlying workspace state under your control. Inspired by Karpathy’s LLM wiki workflow, the system turns connected work data into browsable, compressed memory that both you and the agent can use.
This compression layer preserves the important working facts, relationships, and context from the original data while making memory faster to retrieve and easier for the agent to select from. Instead of manually bringing every detail back into the conversation, your workspace maintains a living memory of what has happened, what matters, and where the agent should look next.
<p align="center">
<img src="docs/images/mem-tree.png" alt="holaOS desktop workspace screenshot" width="1280" />
</p>
# HolaOS vs Other Agents
A high level overview over core dimensions that impacts agent usage:
| | OpenClaw | Hermes Agent | Openhuman | HolaOS |
| --- | --- | --- | --- | --- |
| Agent Position | ✅General Agent | ✅General Agent | ✅Personal Agent | 🔥Working Agent |
| Simple to start | ⚠️ Terminal-first | ⚠️ Terminal-first | ⚠️ Basic UI | ✅Production Grade UI |
| Memory | ⚠️ Plugin-reliant | ⚠️ Self-learning | ✅Memory Tree + Obsidian vault, optional agentmemory backend | 🔥Business level Memory Mechanism: Memory Tree + Semantic Embedding + RAG |
| Integrations | 🚫BYOK | 🚫BYOK | ✅118+ via OAuth | 🔥1000 + via OAuth + Stable for Working |
| Cost | 🚫BYO Model | 🚫BYO Model | ✅Token Juicy | 🔥Per Native Tool token optimizer |
| API sprawl | 🚫BYOK | ⚠️ Multi-vendor | ✅one-account, single account per integration | ✅one-account + multi account per integrations |
| Auto-fetch | 🚫None | 🚫None | ✅20-min sync into memory | ✅30-min sync into memory |
| Native tools | ✅Code-only | ✅Code-only | ✅Code + search + scraper + voice | 🔥Code + Web Search + Browser Use + Wide Search |
| Model selection | 🚫BYOK | 🚫BYOK | ⚠️only one | 🔥one-account all SOTA models |
| Workspace | 🚫None | 🚫None | 🚫None | 🔥Workspace build for Digital Work |
## Table of Contents
- [Quick Start](#quick-start)
- [What you need](#what-you-need)
- [One-Line Install](#one-line-install)
- [Documentation](#documentation)
- [Manual Install](#manual-install)
- [One-Line Agent Setup](#one-line-agent-setup)
- [Contributing](#contributing)
- [OSS Release Notes](#oss-release-notes)
## Quick Start
### One-Line Install
For a fresh-machine bootstrap on macOS, Linux, or WSL, use the repository installer:
```bash
curl -fsSL https://raw.githubusercontent.com/holaboss-ai/holaOS/refs/heads/main/scripts/install.sh | bash -s -- --launch
```
You can also follow the manual path if you want to control each setup step.
## Star the Repository
<p align="center">
<img src="docs/images/star-the-repo.gif" alt="Animated preview from the holaOS star-the-repo video" width="1280" />
</p>
<p align="center"><strong>If holaOS is useful or interesting, a GitHub Star would be greatly appreciated.</strong></p>
## Documentation
All deeper technical and product documentation lives at **[holaos.ai/docs](https://www.holaos.ai/docs)**:
| Section | What's Covered |
| --- | --- |
| [Overview](https://www.holaos.ai/docs/getting-started) | The merged entry page for the environment-engineering thesis and system model |
| [Quick Start](https://www.holaos.ai/docs/getting-started/quick-start) | The fastest path to a working local desktop environment |
| [Workspaces](https://www.What people ask about holaOS
What is holaboss-ai/holaOS?
+
holaboss-ai/holaOS is subagents for the Claude AI ecosystem. Your super agent for work: local-first, learn your working context in mins and never forget it. It has 5.5k GitHub stars and was last updated 10d ago.
How do I install holaOS?
+
You can install holaOS by cloning the repository (https://github.com/holaboss-ai/holaOS) or following the README instructions on GitHub. ClaudeWave also provides quick install blocks on this page.
Is holaboss-ai/holaOS safe to use?
+
Our security agent has analyzed holaboss-ai/holaOS and assigned a Trust Score of 95/100 (tier: Verified). See the full breakdown of passed checks and flags on this page.
Who maintains holaboss-ai/holaOS?
+
holaboss-ai/holaOS is maintained by holaboss-ai. The last recorded GitHub activity is from 10d ago, with 4 open issues.
Are there alternatives to holaOS?
+
Yes. On ClaudeWave you can browse similar subagents at /categories/agents, sorted by popularity or recent activity.
Deploy holaOS to your cloud
Ship this repo to production in minutes. Each platform spins up its own environment with editable env vars.
Maintain this repo? Add a badge to your README
Drop the badge into your GitHub README to show it's tracked on ClaudeWave. Each badge links back to this page and reflects the live Trust Score.
[](https://claudewave.com/repo/holaboss-ai-holaos)<a href="https://claudewave.com/repo/holaboss-ai-holaos"><img src="https://claudewave.com/api/badge/holaboss-ai-holaos" alt="Featured on ClaudeWave: holaboss-ai/holaOS" width="320" height="64" /></a>More Subagents
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
The agent that grows with you
Java 面试 & 后端通用面试指南,覆盖计算机基础、数据库、分布式、高并发、系统设计与 AI 应用开发
Production-ready platform for agentic workflow development.
The agent engineering platform.
🤯 LobeHub is your Chief Agent Operator, organizing your agents into 7×24 operations by hiring, scheduling, and reporting on your entire AI team.