agent-company-ai 0.3.0

Spin up an AI agent company - a business run by AI agents, managed by you

Agent Company AI

Spin up an AI agent company - a business run by AI agents, managed by you.

Agent Company AI lets a solo entrepreneur create a virtual company staffed entirely by AI agents. Each agent has a specific business role (CEO, CTO, Developer, Marketer, etc.), they collaborate on tasks, and you manage everything through a CLI or web dashboard.

Quick Start

pip install agent-company-ai

1. Initialize your company

agent-company-ai init --name "My AI Startup"

You'll be prompted to choose an LLM provider (Anthropic, OpenAI, DeepSeek, Ollama, and more).

2. One-command setup

Spin up a full team from a preset template:

agent-company-ai setup tech_startup --name "Acme AI"
agent-company-ai setup saas --name "CloudCo" --provider anthropic
agent-company-ai setup --list  # see all presets

Available presets: tech_startup (6 agents), agency (6), ecommerce (7), saas (9), consulting (6), content (5), full (10 — all departments)

3. Or hire agents individually

agent-company-ai hire ceo --name Alice
agent-company-ai hire cto --name Bob
agent-company-ai hire developer --name Carol
agent-company-ai hire marketer --name Dave

4. Run autonomously

Give the CEO a goal and watch the company run:

agent-company-ai run "Build a landing page for our new product"

The CEO will break down the goal, delegate tasks to the team, and agents will collaborate to deliver results.

# With limits
agent-company-ai run "Launch MVP" --cycles 3 --timeout 600 --max-tasks 20

Commands

Command	Description
init	Initialize a new company
setup <preset>	Set up a full company from a template
hire <role>	Hire an agent
fire <name>	Remove an agent
team	List all agents
assign "<task>"	Assign a task
tasks	Show the task board
chat <name>	Chat with an agent
run "<goal>"	Autonomous mode
broadcast "<msg>"	Message all agents
dashboard	Launch web dashboard
status	Company overview
roles	List available roles
companies	List all companies in this directory
destroy	Permanently delete a company

Global Options

Flag	Description
--company / -C	Company slug to operate on (default: default)

You can also set the company via environment variable:

export AGENT_COMPANY_NAME=my-startup
agent-company-ai team  # operates on my-startup

Multi-Company Support

Run multiple independent companies in the same directory. Each company gets its own config, database, and agents:

# Default company
agent-company-ai init --name "Acme AI"
agent-company-ai hire ceo --name Alice

# Create a second company
agent-company-ai -C my-startup init --name "My Startup" --provider anthropic
agent-company-ai -C my-startup hire ceo --name Bob

# List all companies
agent-company-ai companies

# Destroy a company
agent-company-ai destroy --company my-startup
agent-company-ai destroy --yes  # destroy default, skip confirmation

Directory layout:

.agent-company-ai/
    default/
        config.yaml
        company.db
    my-startup/
        config.yaml
        company.db

Existing single-company setups are automatically migrated into default/ on first access.

Available Roles

Role	Title	Reports To
ceo	Chief Executive Officer	Owner
cto	Chief Technology Officer	CEO
developer	Software Developer	CTO
marketer	Head of Marketing	CEO
sales	Head of Sales	CEO
support	Customer Support Lead	CEO
finance	CFO / Finance	CEO
hr	Head of HR	CEO
project_manager	Project Manager	CEO

LLM Providers

Supports 11 providers out of the box:

Provider	Models	API Key Env Var
Anthropic (default)	Claude Sonnet 4.5, etc.	ANTHROPIC_API_KEY
OpenAI	GPT-4o, etc.	OPENAI_API_KEY
DeepSeek	DeepSeek-R1	DEEPSEEK_API_KEY
MiMo	MiMo-7B-RL	DEEPSEEK_API_KEY
Kimi	Kimi-K2	MOONSHOT_API_KEY
Qwen	Qwen-Max	DASHSCOPE_API_KEY
MiniMax	MiniMax-M1	MINIMAX_API_KEY
Ollama	Llama 3.1, etc.	None (local)
Together	Llama, Mixtral, etc.	TOGETHER_API_KEY
Groq	Fast open-source inference	GROQ_API_KEY
OpenAI-compatible	Any endpoint	Custom

Configure different providers per agent:

llm:
  default_provider: anthropic
  anthropic:
    api_key: ${ANTHROPIC_API_KEY}
    model: claude-sonnet-4-5-20250929
  openai:
    api_key: ${OPENAI_API_KEY}
    model: gpt-4o
    base_url: https://api.openai.com/v1  # or any compatible endpoint

agents:
  - name: Alice
    role: ceo
    provider: anthropic
  - name: Bob
    role: developer
    provider: openai

Web Dashboard

agent-company-ai dashboard --port 8420

Features:

• Org Chart - Visual company hierarchy
• Agent Roster - See all agents and their roles
• Task Board - Kanban-style task management
• Chat - Talk directly to any agent
• Activity Feed - Real-time event stream via WebSocket
• Autonomous Mode - Set goals and monitor progress from the UI
• Cost Tracker - Real-time API cost breakdown by agent and model

Built-in Agent Tools

Agents have access to these tools based on their role:

Tool	Description
web_search	Search the web via DuckDuckGo (no API key needed)
read_file / write_file	File operations in the workspace (sandboxed)
code_exec	Execute Python code (restricted builtins)
shell	Run shell commands (30s timeout, dangerous patterns blocked)
delegate_task	Delegate work to other agents
report_result	Submit task results

Autonomous Mode

The company runs in CEO-driven cycles:

1. Plan - CEO breaks the goal into tasks and delegates to the team
2. Execute - Agents work on tasks in parallel waves
3. Review - CEO evaluates progress and decides: DONE, CONTINUE, or FAILED
4. Loop - Repeat until goal achieved or limits reached

Configurable limits (in config.yaml or via CLI flags):

• max_cycles: 5 — CEO review loops
• max_waves_per_cycle: 10 — parallel execution waves per cycle
• max_total_tasks: 50 — hard cap on tasks
• max_time_seconds: 3600 — wall-clock timeout
• max_cost_usd: 0.0 — spending cap (0 = unlimited)

Custom Roles

Create custom roles by adding YAML files:

# .agent-company-ai/roles/custom_analyst.yaml
name: analyst
title: "Data Analyst"
description: "Analyzes data and creates reports"
system_prompt: |
  You are a data analyst at {company_name}.
  Your expertise: data analysis, visualization, reporting.
  Team: {team_members}
  Delegates: {delegates}
default_tools:
  - code_exec
  - file_io
can_delegate_to: []
reports_to: cto

Enterprise Customization & Consulting

"We build AI agent workforce for your company"

• Implementation fee: $10k–100k+
• Ongoing support: $2k–10k/month

Please contact [e-mail]

License

MIT