← 返回命令列表

Linux command

paseo 命令

文件

复制后可按需替换文件名、目录或参数。

常用示例

Start the local daemon

paseo daemon start

Launch an agent

paseo run --provider [claude/opus-4.6] "[implement user authentication]"

Launch an agent in its own git worktree

paseo run --worktree [feature-branch] --provider [codex] "[refactor the payment module]"

List currently running agents

paseo ls

Stream live output

paseo attach [agent_id]

Send a follow-up instruction

paseo send [agent_id] "[also add unit tests]"

Target a remote daemon

paseo run --host [host:port] --provider [opencode] "[review the diff]"

说明

paseo is a command-line client and local daemon that orchestrates multiple AI coding agents behind a single interface. Each invocation of paseo run spawns an agent process under one of the supported providers (Claude Code, Codex, GitHub Copilot, OpenCode, Pi) and registers it with the daemon, which keeps the agent alive across terminal sessions and exposes it for follow-up messages or live attachment. A typical workflow is to start the daemon once with paseo daemon start, fire off several agents in parallel on different prompts or branches, list them with paseo ls to see status and IDs, and then attach or send to interact with whichever one needs supervision. Because each agent can be pinned to its own git worktree, parallel agents work on isolated branches without stomping on each other's edits. The daemon model also enables cross-device usage: a daemon running on a workstation can be reached from a laptop by passing --host, so long-running agent sessions outlive any single shell or device. Paseo positions itself as a thin coordination layer above the underlying coding-agent CLIs rather than a separate AI: model selection and prompts are handed straight to the chosen provider.

参数

--provider _PROVIDER_
Required for paseo run. Selects the underlying coding agent, e.g. claude/opus-4.6, codex, copilot, opencode, pi. Use the provider's own model identifier where applicable.
--worktree _BRANCH_
Run the agent inside a dedicated git worktree on _BRANCH_, so its file edits stay isolated from other parallel agents.
--host _HOST_
Route the command to a remote Paseo daemon at _HOST_ (e.g. host:port) instead of the local one.

FAQ

What is the paseo command used for?

paseo is a command-line client and local daemon that orchestrates multiple AI coding agents behind a single interface. Each invocation of paseo run spawns an agent process under one of the supported providers (Claude Code, Codex, GitHub Copilot, OpenCode, Pi) and registers it with the daemon, which keeps the agent alive across terminal sessions and exposes it for follow-up messages or live attachment. A typical workflow is to start the daemon once with paseo daemon start, fire off several agents in parallel on different prompts or branches, list them with paseo ls to see status and IDs, and then attach or send to interact with whichever one needs supervision. Because each agent can be pinned to its own git worktree, parallel agents work on isolated branches without stomping on each other's edits. The daemon model also enables cross-device usage: a daemon running on a workstation can be reached from a laptop by passing --host, so long-running agent sessions outlive any single shell or device. Paseo positions itself as a thin coordination layer above the underlying coding-agent CLIs rather than a separate AI: model selection and prompts are handed straight to the chosen provider.

How do I run a basic paseo example?

Run `paseo daemon start` in a terminal, then adjust file names, paths, flags, or remote targets for your system.

What does --provider _PROVIDER_ do in paseo?

Required for paseo run. Selects the underlying coding agent, e.g. claude/opus-4.6, codex, copilot, opencode, pi. Use the provider's own model identifier where applicable.