nex-crm/wuphf

What it solves

WUPHF creates a collaborative "office" for AI agents, allowing multiple specialized AI employees (such as a CEO, PM, engineers, and designers) to work together in a shared space. It solves the problem of AI agents disappearing behind an API by making their work, arguments, and task claims visible and transparent to the user.

How it works

WUPHF operates as a shared office environment where agents are push-driven and start fresh sessions per turn to maintain efficiency and reduce token costs. It integrates with various agent runtimes like Claude Code, Codex, and Hermes Agent.

Key components include:

• Shared Memory: A git-native markdown wiki that serves as a living knowledge graph for the team, alongside individual agent notebooks for private context.
• Broker System: A push-driven architecture that wakes agents only when needed, eliminating idle token burn.
• Action Providers: Agents can take real-world actions via a local CLI (One CLI) or cloud-hosted SaaS integrations (Composio).
• Bridges: The system can bridge to Telegram or OpenClaw agents to bring external AI entities into the shared office.

Who it’s for

It is designed for developers and teams who want to deploy a coordinated team of AI agents to handle work 24/7, while maintaining full visibility into their processes and a shared knowledge base.

Highlights

• Collaborative AI Office: A visible workspace where multiple AI roles (CEO, PM, etc.) collaborate and and ship work.
• Git-native Memory: Uses a local markdown wiki for shared knowledge and per-agent notebooks for private notes.
• Prompt Efficiency: Employs fresh sessions and per-role tool scoping to maximize prompt caching and reduce costs.
• Zero Idle Burn: Agents are activated only via push notifications, avoiding constant polling.
• External Integrations: Supports bridging to Telegram and OpenClaw, and supports various LLM providers including Ollama and Claude Code.