WUPHF by Nex.ai

WUPHF by Nex.ai 介绍

WUPHF is an open-source, self-hosted platform that creates a collaborative office of AI agents (like Claude Code, Codex, OpenClaw) with a shared, persistent knowledge base. It allows agents to autonomously execute tasks and hand off work while maintaining context, reducing token waste and manual coordination. The system runs locally with your own API keys and supports a TUI, web UI, and Telegram interface.

核心功能

Shared git-backed wiki for persistent context across agents

Supports multiple AI agent types (Claude Code, Codex, OpenClaw) and local LLMs

Fresh sessions per turn with 97% prompt cache hit rates, minimizing token usage

Push-driven agent wakes with zero idle burn, reducing costs

Open source (MIT), self-hosted, no telemetry or data leaving your machine

典型使用场景

Software teams orchestrating multiple AI coding assistants (e.g., engineer, PM, marketing) who collaborate on complex projects

Individual developers running multiple agents for code generation, review, and deployment without context loss

Startups automating repetitive tasks like development, content creation, and analysis with a single, local environment

为什么适合初创团队

Startups need WUPHF because it eliminates repetitive context-pasting across multiple AI agents, saving tokens and time. Its open-source, self-hosted nature means no per-seat costs or data leakage, making it ideal for bootstrapped teams. Plus, it scales from a single developer to a full team of AI employees without cloud dependencies.

常见问题

How does WUPHF keep context between agents?

All agents share a git-backed wiki that stores typed fact graphs and per-agent notebooks. Context is written and read from this shared brain, so agents never lose track of tasks.

Is WUPHF free to use?

Yes, WUPHF is fully open source under the MIT license. You can install it with `npx wuphf@latest` and run it locally with your own API keys.

Does my data leave my machine?

No. The runtime and context are local; the only network calls are to your configured LLM provider for inference. If you use a local model, nothing leaves your machine.

What AI agents does WUPHF support?

It supports Claude Code, Codex, OpenClaw, and any local LLM via OpenCode. You can add or swap agents as needed.

How do I install and run WUPHF?

Simply run `npx wuphf@latest` in your terminal, or clone the repo from GitHub and build with Go. The web UI opens at localhost:7891.