cli-anything-agentscope

name	cli-anything-agentscope
description	CLI harness for AgentScope Demo — agent chat, session management, knowledge base, file operations, and HITL approval

cli-anything-agentscope

Command-line interface for the AgentScope Demo server. Provides agent chat, session management, knowledge base, file upload/download operations, and human-in-the-loop (HITL) approval flows.

Prerequisites

• AgentScope Demo server running (mvn spring-boot:run on http://localhost:8080)
• Python 3.10+
• pip install cli-anything-agentscope

Installation

pip install -e .
# or from PyPI:
pip install cli-anything-agentscope

Global Options

Flag	Description
--json	Output all results as JSON (for agent consumption)
--server URL	Override server URL (default: http://localhost:8080)
AGENTSCOPE_BASE_URL	Env var for server URL

Command Groups

server — Server connection and status

Command	Description
server status	Check server health and connectivity
server info	Show detailed server information

agent — Agent listing and configuration

Command	Description
agent list	List all available agents
agent info <agent_id>	Show detailed agent configuration
agent messages <agent_id>	List chat history for an agent
agent sample-prompt <agent_id> <index>	Get sample prompt by index
agent skill-info <skill_name>	Show skill details
agent tool-info <tool_name>	Show tool details

session — Session management

Command	Description
session list	List all sessions
session create [-a AGENT_ID]	Create new session
session delete <session_id>	Delete a session
session use <session_id>	Set active session for REPL

chat — Send messages to agents

Command	Description
chat send <message> [-a AGENT_ID] [-s SESSION_ID] [--file PATH] [--stream/--no-stream]	Send message
chat metrics <message>	Send message and show debug metrics
chat approve <approval_id> [--reject] [--reason TEXT]	Approve/reject HITL request

knowledge — Knowledge base management

Command	Description
knowledge list	List indexed documents
knowledge status	Show indexing status
knowledge upload <file_path>	Upload document to knowledge base
knowledge remove <file_name>	Remove document from knowledge base
knowledge search <query> [-n LIMIT] [-t THRESHOLD]	Search knowledge base

upload — File upload and download

Command	Description
upload file <file_path>	Upload file to server
upload download <file_id> -o <output>	Download file from server

REPL Mode

Running with no arguments enters interactive REPL mode:

$ cli-anything-agentscope

REPL commands:

• chat <message> — Send message (streaming)
• chat! <message> — Send message with metrics
• approve <approval_id> [--reject] — Approve/reject HITL request
• agent list|use <id>|info <id>|messages <id>|skill-info <name>|tool-info <name>
• session list|create [agent]|use <id>|delete <id>
• knowledge list|upload <path>|search <query>|status
• upload <path>
• server status
• help — Show available commands
• quit / exit — Exit REPL

Usage Examples

One-shot commands (JSON output)

# Check server status
cli-anything-agentscope --json server status

# List all agents
cli-anything-agentscope --json agent list

# Send a chat message
cli-anything-agentscope --json chat send --no-stream "What is 2+2?"

# Create a session and send messages
cli-anything-agentscope --json session create -a chat-basic
cli-anything-agentscope --json chat send -s <session_id> "Hello"

# Upload and search knowledge base
cli-anything-agentscope --json knowledge upload report.pdf
cli-anything-agentscope --json knowledge search "quarterly revenue"

# Check knowledge indexing status
cli-anything-agentscope --json knowledge status

# Upload a file
cli-anything-agentscope --json upload file document.docx

# Get agent chat history
cli-anything-agentscope --json agent messages chat-basic

# Get skill/tool details
cli-anything-agentscope --json agent skill-info docx
cli-anything-agentscope --json agent tool-info parse_docx

Agent-specific usage patterns

# Chat with a specific agent
cli-anything-agentscope chat send -a tool-test-simple "What time is it?"

# Use vision agent with an image
cli-anything-agentscope chat send -a vision-analyzer --file photo.jpg "Describe this image"

# Multi-agent pipeline (sequential)
cli-anything-agentscope --json agent info doc-analysis-pipeline

# Routing agent
cli-anything-agentscope --json agent info smart-router

# Loop pattern (write-review-revise)
cli-anything-agentscope chat send -a copywriter-refiner "Write a product description"

# HITL approval
cli-anything-agentscope chat approve <approval_id>
cli-anything-agentscope chat approve <approval_id> --reject --reason "Unsafe operation"

JSON Output Format

All commands support --json for machine-readable output:

// server status
{"status": "ok", "url": "http://localhost:8080", "agent_count": 30}

// agent list
[{"agentId": "chat-basic", "name": "Basic Chat", "type": "SINGLE", ...}]

// chat send --no-stream
{"text": "response text", "events": [...]}

// knowledge search
{"query": "...", "count": 2, "results": [{"content": "...", "score": "0.85"}]}

// chat approve
{"text": "response after approval", "events": [...]}

Error Handling

• Server unreachable: Clear error message with start instructions
• HTTP errors: Status code and message included in output
• All errors exit with code 1 in one-shot mode