Skip to main content

runanywhere

Set up and use RunAnywhere from the terminal — install rcli, sign in, pick a coding harness, run a model, check spend. Use when the user wants to get started with RunAnywhere, run a harness like opencode against a hosted or on-device model, or asks what their usage is.

来源信息

仓库
RunanywhereAI/RCLI
最近来源活动
2026年9月4日 08:29
检测到的 SKILL.md 语言
英语
星标
1,544
分支
85

安装方式

默认使用会先检查来源的 Prompt;你也可以切换为直接命令,或下载本地副本。

检查来源文件

决定是否安装前,请先阅读 SKILL.md,以及 SkillsMP 当前展示的配套文件。

正在显示 SKILL.md

SKILL.md
来源说明 · 只读预览
name
runanywhere
description
Set up and use RunAnywhere from the terminal — install rcli, sign in, pick a coding harness, run a model, check spend. Use when the user wants to get started with RunAnywhere, run a harness like opencode against a hosted or on-device model, or asks what their usage is.
# RunAnywhere `rcli` is one CLI for two things: models running **on this machine**, and models served from **RunAnywhere Cloud**. The same commands cover both — if a model is on the machine it is served locally, otherwise the request goes to the console the user is signed in to and is metered against their balance. ## First check what is already true ```bash rcli whoami # signed in? which console? rcli backends # which engines this build linked ``` `rcli whoami` failing with "not signed in" is the only thing that needs fixing before anything else works against the cloud. On-device models need no account. ## Signing in ```bash rcli login ``` Opens the console in a browser. The person signs in with Google or GitHub, approves the terminal, and the CLI stores a key in `~/.config/rcli`. There is no password and no organization step. If a browser cannot open, `rcli login --no-browser` prints the URL to visit. ## Coding harnesses A harness is an existing coding tool that `rcli` wires to a model. Today that is **opencode**. ```bash rcli opencode --cloud -m glm-5.3 # hosted, metered rcli opencode -m qwen3-0.6b # a model on this machine ``` If opencode is not installed, `rcli` says so and prints the install command (`npm i -g opencode-ai`) rather than failing. Install it, then run the same line again. Offer to explain a harness before running it. Someone who has just signed up does not yet know what opencode is, and "would you like me to explain how the opencode harness works?" is a better second message than a launched TUI. ## Running a model directly ```bash rcli pull qwen3-0.6b # download it rcli list # what is downloaded rcli run qwen3-0.6b # talk to it ``` Models land in `~/.local/share/runanywhere`. Nothing is downloaded until asked. ## Spend ```bash rcli usage # credit left, then input/output/cache tokens and spend rcli usage --json ``` Read-only, and scoped to the signed-in account. ## When something is wrong - **"not signed in"** — `rcli login`. - **"that key is not valid"** — the key was revoked or expired; `rcli login` again. - **opencode not installed** — `npm i -g opencode-ai`. - **a model is slow or unavailable** — `rcli backends` shows which engines this build actually linked; a model needing an engine that is not there will not run. ## What not to do Do not print, log or echo the contents of `~/.config/rcli/credentials.json`. It holds a key with the person's credit behind it.
在 GitHub 查看