Project Golem AI Work Hub
Bring AI conversations, tools, and long-term memory together in a self-hosted workbench.
- Source repo
- Arvincreator/project-golem
- Stars
- ★ 642
- Last updated
- 5mo ago
- License
- NOASSERTION
- Primary language
- JavaScript
- FA score
- 46/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platform
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A solo user wants one local Dashboard for AI chat, Skills, MCP tools, and memory.
- Not a fit if
- Users who need a hosted service with no local installation
- Teams unable to run Node.js 20 through 22
- Maintenance
- Commits in the last 6 monthslast commit 2026-06-22latest release v9.6.14 (2026-05-18)
- Source review
- 46/100 · Major gaps
What does this agent do, and when should you use it?
Project Golem is a Node.js AI work system that brings model conversations, tool actions, and memory management into a Web Dashboard. It supports Gemini Web through Playwright, Ollama, and LM Studio, with backend selection controlled by configuration. The Dashboard includes chat, Skills and MCP management, multi-agent discussions, memory retrieval, and features such as a calendar, stock board, and text RPG. Run the runtime and Dashboard with npm run dashboard, or deploy with Docker Compose; the README recommends running Node.js directly for personal use. This is a system to install and operate yourself; the README does not describe a vendor-hosted service.
After starting the Project Golem runtime, users can interact with the selected model through the Dashboard or chat commands. The system supports Gemini Web, Ollama, and LM Studio backends, and extends workflows through MCP servers and Skills; its long-term memory offers lancedb-pro and native modes with summarization and retrieval. Users can create multi-agent role groups to discuss a topic and consolidate consensus, and manage calendar events, stock watchlists, RPG interactions, and relationship diaries in the Dashboard. Optional Telegram and Discord bridges provide external channels. High-risk actions have an Action Gate approval flow, alongside remote-access protection settings.
- A solo user wants one local Dashboard for AI chat, Skills, MCP tools, and memory.
- Someone working on a long-running project wants memory and summaries to carry context across conversations.
- Product, engineering, security, or operations roles need a structured discussion with role-based agents and a consolidated outcome.
- Users who prefer local or private models want to select Ollama or LM Studio as the backend.
- Users who want planning and market information in their AI workflow can use the collaboration calendar and Taiwan or US stock board.
- Users who want to interact through Telegram or Discord can configure the corresponding optional bridge.
How do you install or deploy this agent?
The documented runtime requirements are Node.js >=20 <23 and npm. Gemini Web mode also requires Chromium or Google Chrome. The recommended setup is to run setup.sh from the project directory; use Git Bash on Windows and Terminal on macOS. Docker Compose is also documented for server or VPS deployment.
How do you use this agent?
After installation, use one of these commands to start the runtime with the Web Dashboard, or to start only the core runtime. The Dashboard defaults to http://localhost:3000/dashboard. Before the first session, choose a backend and configure its endpoint or model settings; the README provides examples for Gemini Web, Ollama, and LM Studio. Remote access is disabled by default. If you enable it, the README recommends setting a password and system operation token and using a firewall or VPN.
What are this agent's strengths and limitations?
- Offers configurable Gemini Web, Ollama, and LM Studio backends, including local and private-model options.
- Combines chat, memory, Skills, MCP, settings, and multi-agent collaboration in one Dashboard.
- Provides
lancedb-proandnativelong-term memory modes with summarization and retrieval. - Documents Docker Compose and headless/noVNC deployment options for self-hosting.
- Requires users to install and operate a Node.js runtime; the documented range is Node.js 20 through 22.
- Gemini Web mode requires Chromium or Google Chrome and uses Playwright for browser interaction.
- The README does not state model or service pricing or give complete minimum hardware requirements.
- The README describes the license as Source-Available Non-Commercial; commercial users need to review the license terms.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Project Golem AI Work Hub This agent | 46 · Major gaps | Web app | ★ 642 | 5mo ago | JavaScript | — |
| Switch AI Team Workspace | 70 · Some gaps | Desktop app | ★ 4.7k | today | Python | — |
| Cherry Studio | 51 · Major gaps | Desktop appFreemium | ★ 54k | today | TypeScript | OpenAI API · Claude API |
| Markus AI Workforce | 51 · Major gaps | Desktop appFree + model costs | ★ 219 | today | TypeScript | OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
The README says remote access is off by default and describes passwords, an operation token, and an Action Gate for high-risk operations; configuration also lists Telegram and Discord authentication modes. These provide some control evidence, but the packet does not explain permission scopes, data flows, or sensitive-memory handling. The Chrome DevTools skill grants full browser control and instructs script-based clicking and data collection without describing granular authorization or confirmation for external actions, so points are deducted. Backup and restore are claimed, but the sample does not show a concrete recovery path. An author name and license holder are listed, but publisher identity is unverified and maintenance or update commitments are unclear.
The README, package.json, and runtime source support basic Node 20–22 requirements and npm commands for startup, tests, and architecture checks. CI declares installation, architecture checks, tests, and a high-severity production dependency audit. However, version information conflicts: the README says 9.7.9, the runtime entry comment says v9.5, and the package description also says v9.5. Error logging and health-check commands are present, but the sample does not establish recovery behavior when dependencies fail or give actionable user guidance for failures, so only partial credit is warranted.
The README names personal workspace, local models, long-running projects, multi-agent decisions, and tool experiments as scenarios, and lists Gemini Web, Ollama, LM Studio, Node, and Chromium requirements. The skills give explicit triggers for scheduling, persona changes, and browser tasks. This supports ordinary scenario and environment fit. However, the advertised capabilities span browser control, files, external services, social bridges, and autonomous scheduling, while the sample does not clearly define permission boundaries or limits for those capabilities.
The README provides a project map, documentation index, installation steps, common commands, configuration examples, and use cases. The license clearly states non-commercial terms, commercial-permission requirements, and third-party component terms, earning stronger scores on those items. Deductions reflect inconsistent product/version naming across the README, package.json, and runtime comment, no visible changelog, no clear known-limitations list, and no clear maintenance responsibility or update path. The metadata says NOASSERTION, but the repository license text itself is available for review.
The README organizes chat, memory, skills, MCP, multi-agent work, and dashboard capabilities into recognizable workflows and provides commands and configuration examples. Long-term memory and local-model support have understandable value for sustained work. Cost-benefit evidence is limited: the product requires Node and npm, Chromium for Gemini Web, and multiple external services and broad runtime capabilities; the packet does not quantify usage costs, maintenance burden, or tradeoffs.
Some README claims can be traced to configuration, runtime entry code, skill files, package.json, and declared CI steps, including backend choices, scheduling, browser tools, and test commands. Cross-source corroboration remains limited: the README claims Action Gate, memory management, and backup/restore features, but the supplied sample lacks their implementations; CI shows declared test steps, not results. Documentation sometimes presents capabilities as established without consistently distinguishing product claims from implementation directly visible in the sample, so full marks are not justified.
- The product spans browser automation, external services, social bridges, and autonomous scheduling; before deployment, verify each integration's permissions, data destinations, and confirmation path for high-risk actions.
- Version information conflicts across the README, package.json, and runtime comment. The license is non-commercial; commercial use requires prior written permission.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
V9.6.6 這次變更了什麼 1. 版本更新 - package.json、README.md、docs/README.en.md、Dashboard 版本字串更新到 9.6.6 2. MCP 啟動設定調整 - data/mcp-servers.json 更新 3. Action/技能執行流程調整 - NeuroShunter、SkillHandler、ActionExecutionGate 有修正 - skillsConfig 與多個 skill 說明/模組同步調整 4. 參考文件流程與 Dashboard 串接 - ReferenceFileService、api.reference-files、reference-files 頁面與 console 顯示有更新 5. 舊的 schedule/list-schedules 模組清理 - src/skills/core/* 與 src/skills/modules/* 中對應檔案已移除 --- 這次改進了什麼(效果) - 工具/技能 action 路由判斷更一致,減少被誤判或落錯 lane 的情況 - 參考文件讀取與前端可視化更完整,問題排查更直觀 - 技能配置更精簡(移除舊模組),降低重複邏輯與維護成本 …View on GitHub ↗
FAQ
Does Project Golem cost money to use?
Can I run it entirely on my own machine?
What do I need to install?
>=20 <23 and npm. Gemini Web mode also needs Chromium or Google Chrome.Can I use it from Telegram or Discord?
Is it suitable for commercial use?
Is Project Golem AI Work Hub open source?
Is Project Golem AI Work Hub safe to use?
Is Project Golem AI Work Hub still maintained?
Who makes Project Golem AI Work Hub?
Thanks — we use this to improve the page.