Multi-Agent CLI

Define and run multiple agents with custom roles, tools, and instructions from the command line

The praisonai agents command allows you to define and run multiple agents directly from the command line without creating a YAML configuration file.

Quick Start

# Run a single agent with tools
praisonai agents run \
  --agent "researcher:Research Analyst:internet_search" \
  --task "Find the latest AI news"

# Run multiple agents
praisonai agents run \
  --agent "researcher:Research Analyst:internet_search" \
  --agent "writer:Content Writer:write_file" \
  --task "Research AI trends and write a summary report"

Commands

Run Agents

Execute one or more agents with a specified task.

praisonai agents run [options]

Required Options:

Option	Description
--agent, -a	Agent definition (can be used multiple times)
--task, -t	Task for the agents to complete

Optional Options:

Option	Description
--process, -p	Execution process: sequential (default) or parallel
--llm, -m	LLM model to use for all agents
--instructions, -i	Additional instructions for all agents
--verbose, -v	Enable verbose output

List Available Tools

List all tools that can be assigned to agents.

praisonai agents list

Output:

Available tools for agents:
------------------------------
  internet_search: Search the web
  read_file: Read file contents
  write_file: Write to a file
  list_files: List directory contents
  execute_command: Execute shell commands
  read_csv: Read CSV files
  write_csv: Write CSV files
  analyze_csv: Analyze CSV data

Agent Definition Format

Agents are defined using a colon-separated format:

name:role:tools:goal

Part	Required	Description
name	Yes	Unique identifier for the agent
role	Yes	The agent’s role/title
tools	No	Comma-separated list of tools
goal	No	Specific goal for the agent

Examples

# Basic agent with one tool
--agent "researcher:Research Analyst:internet_search"

# Agent with multiple tools
--agent "analyst:Data Analyst:read_csv,write_csv,analyze_csv"

# Agent without tools
--agent "helper:Assistant"

# Agent with goal
--agent "writer:Writer:write_file:Create high-quality content"

Usage Examples

Single Agent

# Simple question
praisonai agents run \
  --agent "assistant:Helper" \
  --task "What is the capital of France?"

# With web search
praisonai agents run \
  --agent "researcher:Researcher:internet_search" \
  --task "Find the latest news about AI"

Multiple Agents (Sequential)

Agents execute one after another, passing context between them.

praisonai agents run \
  --agent "researcher:Research Analyst:internet_search" \
  --agent "writer:Content Writer:write_file" \
  --task "Research renewable energy trends and write a blog post"

Multiple Agents (Parallel)

Agents execute simultaneously for faster results.

praisonai agents run \
  --agent "analyst1:Market Analyst:internet_search" \
  --agent "analyst2:Tech Analyst:internet_search" \
  --process parallel \
  --task "Analyze the current state of AI industry"

With Custom LLM

praisonai agents run \
  --agent "coder:Developer:execute_command" \
  --llm "gpt-4o" \
  --task "Write a Python script to calculate fibonacci numbers"

With Additional Instructions

praisonai agents run \
  --agent "analyst:Data Analyst:read_csv,analyze_csv" \
  --instructions "Be thorough and provide detailed analysis" \
  --task "Analyze the sales data in data.csv"

Data Analysis Pipeline

praisonai agents run \
  --agent "reader:Data Reader:read_csv" \
  --agent "analyst:Data Analyst:analyze_csv" \
  --agent "reporter:Report Writer:write_file" \
  --task "Read sales.csv, analyze trends, and write a report"

Code Development

praisonai agents run \
  --agent "architect:Software Architect" \
  --agent "developer:Developer:execute_command,write_file" \
  --agent "tester:QA Engineer:execute_command" \
  --task "Design and implement a REST API for user management"

Available Tools

Tool	Description
internet_search	Search the web for information
read_file	Read contents of a file
write_file	Write content to a file
list_files	List files in a directory
execute_command	Execute shell commands
read_csv	Read and parse CSV files
write_csv	Write data to CSV files
analyze_csv	Analyze CSV data with statistics

Execution Modes

Sequential (Default)

Agents execute one after another. Each agent receives the context from previous agents.

praisonai agents run \
  --agent "agent1:Role1:tool1" \
  --agent "agent2:Role2:tool2" \
  --process sequential \
  --task "Complete this task"

Parallel

Agents execute simultaneously. Useful for independent tasks.

praisonai agents run \
  --agent "agent1:Role1:tool1" \
  --agent "agent2:Role2:tool2" \
  --process parallel \
  --task "Complete this task"

Output

The command displays:

1. Agent information (name, role)
2. Task description
3. Agent responses
4. Final result

Example Output:

🚀 Running 2 agents (sequential mode)...
  - researcher: Research Analyst
  - writer: Content Writer

╭─ Agent Info ─────────────────────────────────────────────────────────────────╮
│  👤 Agent: researcher                                                        │
│  Role: Research Analyst                                                      │
╰──────────────────────────────────────────────────────────────────────────────╯

╭────────────────────────────────── Response ──────────────────────────────────╮
│ [Agent's response here]                                                      │
╰──────────────────────────────────────────────────────────────────────────────╯

==================================================
RESULT:
==================================================
[Final result here]

Comparison with agents.yaml

Feature	praisonai agents run	agents.yaml
Setup	No file needed	Requires YAML file
Complexity	Simple, quick tasks	Complex workflows
Reusability	One-time use	Reusable configuration
Customization	Basic options	Full configuration
Best for	Quick experiments	Production workflows

Tips

1. Start Simple: Begin with a single agent, then add more as needed
2. Use Descriptive Roles: Clear roles help agents understand their purpose
3. Match Tools to Tasks: Only assign tools that are relevant to the task
4. Sequential for Dependencies: Use sequential mode when agents depend on each other
5. Parallel for Speed: Use parallel mode for independent tasks

Troubleshooting

Agent Not Using Tools

Ensure the tool name is spelled correctly and the tool is available:

praisonai agents list

Task Not Completing

Try adding more specific instructions:

praisonai agents run \
  --agent "helper:Assistant" \
  --instructions "Be concise and direct" \
  --task "Your task here"

Model Errors

Specify a different model:

praisonai agents run \
  --agent "helper:Assistant" \
  --llm "gpt-4o-mini" \
  --task "Your task here"