English | 简体中文

Command-line tool for managing AI-agent infrastructure on the AgentRun platform.

ar (or agentrun) is a single-binary CLI that wraps the AgentRun Python SDK. It lets developers, CI pipelines, and LLM-powered agents create and operate sandboxes, tools, skills, model services and — most importantly — super agents: platform-hosted AI agents that you configure declaratively without writing or deploying any runtime code.

• One-command super agent — ar super-agent run creates a hosted agent and drops you into a chat REPL in seconds.
• Declarative deployment — Kubernetes-style YAML (ar sa apply -f superagent.yaml) for reproducible, version-controlled agents.
• Six resource groups — config, model, sandbox, tool, skill, super-agent, all following the same ar <group> <action> pattern.
• Multi-profile config — store multiple sets of credentials in ~/.agentrun/config.json and switch with --profile.
• Multiple output formats — json (default), table, yaml, and quiet for shell piping.
• Agent-friendly — JSON-by-default output, deterministic exit codes, no interactive prompts when stdin isn't a TTY.
• Rich sandbox primitives — code execution, file system, process management, and CDP/VNC-backed browser automation.
• Single-file distribution — PyInstaller produces standalone ar / agentrun binaries for Linux, macOS and Windows (x86_64 + arm64).

Download a single self-contained binary from Releases. No Python required.

Linux / macOS (x86_64 or arm64):

curl -fsSL https://raw.githubusercontent.com/Serverless-Devs/agentrun-cli/main/scripts/install.sh | sh

Windows (x86_64, PowerShell):

irm https://raw.githubusercontent.com/Serverless-Devs/agentrun-cli/main/scripts/install.ps1 | iex

Pin a specific version with AGENTRUN_VERSION=v0.1.0 …. Change the install directory with AGENTRUN_INSTALL=…. Both installers verify the SHA256 checksum before placing the binary.

Or download the archive manually from the Releases page — naming scheme:

agentrun-<version>-<os>-<arch>.<ext>
# e.g. agentrun-0.1.0-linux-amd64.tar.gz
#      agentrun-0.1.0-darwin-arm64.tar.gz
#      agentrun-0.1.0-windows-amd64.zip

pip install agentrun-cli

git clone https://github.com/Serverless-Devs/agentrun-cli.git
cd agentrun-cli
make install            # editable install into .venv
make build              # standalone binary → dist/agentrun

ar --version            # or: agentrun --version

Two one-time setup steps are required before ar super-agent will work:

AgentRun uses a custom RAM service role — AliyunAgentRunSuperAgentRole — to manage runtime resources on your behalf. Open the link below and confirm in the RAM console:

→ Create AliyunAgentRunSuperAgentRole

Without this role, ar super-agent run / apply will fail at creation time.

The AccessKey you save with ar config set access_key_id ... must belong to a RAM user (or role) that has the AliyunAgentRunFullAccess system policy attached. If you see exit code 3 or AccessDenied, this is almost always the cause.

This CLI covers the QuickStart conversational flow end-to-end. For the full AgentRun experience, head to the Function Compute AgentRun console: https://functionai.console.aliyun.com/cn-hangzhou/agent/

ar config set access_key_id     LTAI5t...
ar config set access_key_secret ***
ar config set account_id        1234567890
ar config set region            cn-hangzhou

Credentials land in ~/.agentrun/config.json under the default profile. Use --profile staging on any command to target a named profile.

$ ar super-agent run --prompt "You are a Python expert"
Creating super agent: super-agent-tmp-20260420213045 ...
Ready. Type your message (/help for commands).

> Write a quicksort in Python
def quicksort(arr):
    if len(arr) <= 1:
        return arr
    pivot = arr[len(arr) // 2]
    left  = [x for x in arr if x < pivot]
    mid   = [x for x in arr if x == pivot]
    right = [x for x in arr if x > pivot]
    return quicksort(left) + mid + quicksort(right)

> /exit
─────────────────────────────────────────────
Super agent created: super-agent-tmp-20260420213045
Last conversation:  conv-9f8e7d6c-xxx
Resume:  ar sa chat super-agent-tmp-20260420213045
Delete:  ar sa delete super-agent-tmp-20260420213045
─────────────────────────────────────────────

The agent persists after you exit, so you can continue the conversation later with ar sa chat <name> — the CLI remembers the last conversation id locally.

Save this to superagent.yaml:

apiVersion: agentrun/v1
kind: SuperAgent
metadata:
  name: my-helper
  description: "My personal assistant"
spec:
  prompt: "You are my helpful assistant"
  tools:
    - mcp-time-sa
  skills: []
  sandboxes: []
  workspaces: []
  subAgents: []

Then deploy it:

ar super-agent apply -f superagent.yaml
# → action: "created"    (first run)
# → action: "updated"    (subsequent runs)

# Chat with it
ar sa chat my-helper

# Single-shot invocation for scripts
ar sa invoke my-helper -m "Plan my day" --text-only

Multi-document YAMLs (--- separated) let you deploy many agents in one call.

• English reference: docs/en/index.md
• 中文手册: docs/zh/index.md

Each page walks through installation, authentication, global options, output formats, exit codes and every command option with runnable examples.

Questions, bug reports and feature requests are welcome on GitHub Issues.

For real-time discussion, join the 函数计算 AgentRun 客户群 on DingTalk — group number 134570017218.

Apache-2.0 — see LICENSE.