wuphf 命令

常用示例

说明

wuphf is a "Slack for AI employees" — a multi-agent collaboration layer that spawns a small office of role-based agents (CEO, PM, engineers, designer, CMO, CRO) sharing a common memory store and a chat-style channel. Instead of running each AI assistant in isolation, wuphf lets them see each other's messages, claim tasks, and ship work together while preserving context across sessions. Under the hood wuphf delegates each agent to an existing CLI such as claude-code, codex, or openclaw, and exposes a web UI by default at http://localhost:7891 with the team roster, a shared channel, and a message composer. A --tui mode renders the same office inside a tmux session for terminal-only environments. Memory is pluggable via --memory-backend: the markdown backend keeps a human-readable Markdown brain on disk, while nex and gbrain integrate with hosted services. Agent rosters are described by the --pack option, and custom packs can be defined in configuration.

参数

--memory-backend _name_

Select the shared-memory backend: nex, gbrain, markdown (default), or none.

--pack _name_

Choose a preconfigured agent team: starter, founding-team, coding-team, lead-gen-agency, or revops.

--provider _name_

LLM CLI to drive each agent: claude-code (default), codex, or openclaw.

--tui

Use the tmux-based terminal UI instead of the web UI.

--no-nex

Disable the Nex backend while keeping local integrations.

--no-open

Do not open the default browser when the web UI starts.

--opus-ceo

Upgrade the CEO agent from Sonnet to Opus.

--collab

Enable collaborative mode (default).

--unsafe

Bypass agent permission checks; intended only for local experimentation.

--web-port _N_

Web UI port (default 7891).

--1o1 _agent_

Start a one-on-one chat with the CEO, or with the named agent if given.

init

Run initial setup and write a configuration file.

shred

Terminate the active office session and clean up local state.