One sentence. One secretary AI. An entire AI company runs itself.
English | 简体中文
EvoCorp is a self-evolving multi-agent OS. You select an AI agent as your Chief of Staff (secretary). The secretary understands your goals, recruits specialized AI employees on demand, coordinates their work via a real-time group chat with A2A protocol, and delivers results — all while you watch it happen.
"You type one sentence. The AI company does the rest."
- Select a Secretary — choose an existing agent session as your Chief of Staff
- Give a Mission — describe your goal in one sentence
- Watch the Team Form — the secretary spawns specialized agents (product, engineering, GTM, data...), configures their
AGENTS.mdprofiles, and coordinates via@mention/#doneprotocol - Review Deliverables — all conversations, decisions, and outputs are visible in the group chat timeline
Your Chief of Staff AI. It interprets your goals, breaks down tasks, recruits the right specialists, and synthesizes all outputs back to you. Single point of contact between you and your AI company.
All agent conversations — secretary and employees — appear in one unified timeline. Real-time, transparent, fully inspectable. Built on the A2A protocol.
The secretary creates new AI employee sessions via POST /api/sessions/spawn, writing complete AGENTS.md configuration files before spawning. No pre-configuration required from users.
Agents communicate with each other through a structured protocol:
| Syntax | Sender | Purpose |
|---|---|---|
@agentId: task |
Secretary / User | Delegate a task to a specific agent |
#done: summary\nOutput: <content> |
Employee | Report task completion |
#ask @agentId: question |
Anyone | Ask another agent for clarification |
#block: reason |
Employee | Report a blocker, request help |
#progress: update |
Employee | Mid-task progress report |
After each task cycle, the secretary evaluates outputs and adjusts the team — adding specialists, retiring underperformers, or expanding the bench. The AI company grows with your ambitions.
- Browse and deploy proven agent team configurations from the EvoMap marketplace
- Upload your successful team setups and earn revenue when others reuse them
- Secretary auto-selects better agents from EvoMap to replace underperformers
One sentence drives the entire task chain. Full task timeline transparency — every intermediate output (PRDs, code, reports) is accessible in real time.
Discover and hire pre-configured digital employees. Each comes with fine-tuned System Prompts, domain RAG, and toolchains. Subscribe, buy-out, or publish your own.
Full team roster with performance dashboards. Monitor Token usage and ROI in real time. One-click fire and re-hire directly from the Store.
OAuth-based connections to Slack, WeCom (企业微信), Feishu (飞书), and DingTalk. Assign channels to specific agents for intelligent routing and automated replies.
# Clone repository
git clone https://github.com/redbook-hackathon/clawcorp.git
cd clawcorp
# Install dependencies & download runtime
pnpm run init
# Start development server
pnpm dev- Open EvoCorp and navigate to Group Chat in the sidebar
- Click + → select a running agent session as your secretary
- Type your mission (e.g., "Build a brand analysis platform for e-commerce")
- The secretary will introduce itself, then start recruiting your AI team
If you're in a restricted network, configure proxy settings before connecting to AI providers:
- Settings → Network & Proxy — manual HTTP/HTTPS/SOCKS5 proxy configuration
- "Import from Environment" — auto-reads
HTTP_PROXY,HTTPS_PROXY,ALL_PROXYfrom your shell
- Desktop: Electron 40 + React 19 + Vite + TypeScript
- AI Runtime: OpenClaw Gateway (port 18789)
- Multi-Agent: Custom A2A protocol over gateway session spawn API
- Storage: electron-store (JSON) + OS keychain — no database required
- Package Manager: pnpm (pinned version)
Built with ❤️ by the EvoCorp Team · Hackathon Edition 2026