Hire a team of AI employees. Run them like a real company.
雇一支 AI 员工团队，像经营公司一样使用 AI
English · 简体中文
What is it · Features · Who it's for · Quick start · FAQ · Developers
AgentPulse is an open-source “AI company workspace.” You act like the boss and hire a team of AI employees (a content lead, an operator, sales, support, finance…), assign goals in a group chat, and let them discuss until it's clear before acting. Every task and output is traced, and every high-risk action stops for your sign-off. It isn't another chatbot — it reframes “using AI” as “running a company,” built for solo founders, creators, and indie hackers.
- 🧑💼 Hire in one sentence — Say “I need a content lead for Xiaohongshu” and get an AI employee with a persona, skills, and boundaries. No prompt engineering, nothing to configure.
- 💬 Discuss, then execute — Employees behave like real teammates: they question a fuzzy brief in the group, reach a written consensus brief, and only start once you confirm — never charging ahead blind.
- 📋 Everything is traced — Who's doing what, how far along, and what they produced — the task center shows it all. Conversations scatter; tasks and outputs don't.
- 🛡 The boss holds the keys — Publishing, deploying, spending, sending — every high-risk action is structurally stopped for your approval. The AI can never step past that gate.
- 🌱 Gets better the more you use it — Each employee has long-term memory and distills skills from its work: corrected wording, mistakes it hit — it won't repeat them.
- 🌙 24/7, never off the clock — Employees live on a server: they review data on a schedule and surface topics and ideas on their own, so the company keeps running while you sleep.
- 🔓 Open source, model-agnostic — Source open for study and research; works with any OpenAI-compatible model (DeepSeek by default), and text-only models still handle images / audio / video (auto-converted).
- Solo founders — One person carrying marketing, ops, support and finance — share the load with an AI staff and free yourself up.
- Creators — Topic research, multi-platform copy, performance reviews — handed to an always-on AI operator.
- Indie hackers / small teams — Outsource repetitive business work to “digital employees” and stay on your core product.
- Anyone who wants to use AI well without learning prompts / workflows — The product speaks in companies, employees, group chats, tasks and approvals — not APIs and node graphs.
1. Create your company → get a default staff (secretary, writer, ops, support, finance)
2. State a goal in chat → "Plan next week's content for us"
3. They clarify → "Which theme? How many posts? Any livestream to match?"
4. You confirm the brief → one click turns it into tasks; they start executing
5. You sign off at gates → before publishing, before spending — it goes through you
Desktop workspace (Electron + React), light / dark themes. Below: a real scenario for a Xiaohongshu content studio.
 Task center · tasks auto-created after a confirmed consensus; progress / owner / status all traced |
 Talent market · hire AI employees by role; confirm prompt / skills / MCP before onboarding |
 Idea center · no idle employees — they bank improvements / opportunities / risks / learnings when free |
 Dark theme · the same group discussion + consensus brief, in a dark “cockpit” look |
Requirements: Node.js 20+ · Python 3.12+ · Docker (for PostgreSQL)
git clone git@github.com:Clycheng/agentpulse.git && cd agentpulse
npm install
cd services/api && python3 -m venv .venv && source .venv/bin/activate && pip install -r requirements.txt && cd ../..
docker compose up -d postgres
export AGENTPULSE_DEEPSEEK_API_KEY="your DeepSeek API key"
npm run dev:api # backend
npm run dev:desktop # desktop workspaceOpen the desktop app, register, and you're in your company. More config in docs/.
| Chatbot (ChatGPT-style) | Automation (Dify / n8n-style) | AgentPulse | |
|---|---|---|---|
| Mental model | One window, one assistant | Nodes, triggers, JSON | Run a company |
| Multi-role collaboration | ❌ single assistant | 🟡 you orchestrate | ✅ employees discuss & hand off |
| Process & output kept | ❌ gone when you close it | 🟡 engineer-facing | ✅ tasks / progress / outputs traced |
| Dangerous actions gated | ❌ | 🟡 you wire it | ✅ enforced approvals |
| Learning curve | Low | High | Low (hire & delegate in words) |
How is this different from ChatGPT? ChatGPT is one assistant in one window; close it and the context is gone. AgentPulse is a persistent team — each employee has its own persona, skills and memory, work is traced as tasks, and it runs 24/7 on a server without you keeping a window open.
Can the AI go rogue — spend money or publish on its own? No. Every high-risk action (publishing, deploying, anything that spends money) is structurally stopped for your approval. Irreversible spending (e.g. buying a domain) always stays with you. It's a hard product constraint, not a request to the model.
Do I need to be technical? No. The product speaks in companies, employees, group chats, tasks and approvals — not prompts, workflows or DAGs. If you can run a group chat, you can run AgentPulse.
How is it different from Dify / n8n / Coze? Those are automation / orchestration platforms for people who like to configure — you build nodes, wire triggers, tune prompts. AgentPulse hides all of that behind the “company” metaphor: you hire in natural language, delegate in the group, and click confirm; the multi-agent discussion and division of labor are orchestrated for you.
Which models are supported? Any OpenAI-compatible API, DeepSeek by default. A text-only model still handles images, audio and video (auto-converted by the runtime).
Is it usable now? Alpha: the desktop workspace, group chat, tasks, consensus briefs and the approval gate work; real multi-agent discussion and execution are being wired in. Try it and send feedback.
- ✅ Desktop workspace: group chat, employees, tasks, approvals, consensus briefs
- 🚧 Automatic in-group multi-agent discussion and division of labor
- 🚧 Real execution (coding, publishing, deploying…) behind the approval gate
- 📋 Any role from one sentence · skill market · idea center
- 📋 More scenario templates: sales, support, finance, cross-border e-commerce
See ROADMAP.md.
Stack: Electron + React (desktop) · FastAPI + PostgreSQL (backend) · Hermes Agent (employee runtime) · marketing site in apps/site (static, deploys to Vercel).
- Contributing / picking up as an AI: start with AGENTS.md (the project's north star + conventions)
- Architecture & tech design: docs/tech-design/ · decision records: docs/decisions/
Released under the PolyForm Noncommercial License 1.0.0: free for personal study, research, education, hobby projects, and noncommercial organizations. See LICENSE; commercial use requires a license — see COMMERCIAL.md.
Keywords: AI employees · AI digital employees · AI company workspace · AI workforce · one-person company · hire AI employees · multi-agent orchestration · autonomous agents · ai-agents · multi-agent · ai-employees · ai-workforce · agent-orchestration · solo-founder · indie-hackers · deepseek · open-source-ai · AI 员工 · AI 公司工作台 · 多智能体协作