Getting Started

Start with simply signing up with Agiflow—it's free, so give it a spin.

Simple Workflow

Follow these four steps to start automating your development workflow with AI agents.

1

Sign Up & Create Project

Once you're in, create a "New Project" and click on "Start planning". This sets up your workspace and enables AI agents to understand your project context.

Create New Project

Add a new project to organize your tasks and workflows.

2

Connect Your AI Agent

Launch your AI agent via agent-cli. Connect to the Agiflow backend to sync tasks, or run in standalone mode for local development.

Select Daemon
Choose a daemon to execute your tasks. Connected daemons are ready to use immediately.
Advanced Options
Terminal Mode
Enable terminal mode for direct command-line interaction. Default is chat mode.
Use Docker
Run the agent session in a Docker container for enhanced isolation and security.
3

Run Tasks in Sandbox

Execute tasks safely in isolated sandbox environments. Run multiple concurrent tasks with real-time streaming to your dashboard. Each task runs independently while maintaining full visibility.

Task Dashboard

Implement user authentication

Add JWT-based authentication with login, signup, and password reset functionality

High PriorityAuthentication
Oct 30
John Doe

Fix responsive layout issues

Update CSS media queries and flexbox layouts for mobile compatibility

Medium PriorityUI/UX
Nov 2
Jane Smith

Set up database models

Create PostgreSQL schemas and models for users, projects, and tasks

Low PriorityDatabase
Oct 26
Mike Johnson
4

Review & Iterate

Review completed work through the dashboard. Provide feedback via comments and re-run agents with conversation history to implement improvements seamlessly.

Implement user authentication

TASK-123•High Priority•Updated October 27, 2025 at 11:20 PM

Acceptance Criteria

Create login form with email and password fields
Implement JWT token authentication
Add password strength validation
Include remember me functionality

Add Comment

Comments (2)

John Doe
October 26, 2025 at 11:20 PM

I've completed the basic authentication flow. The login form and JWT implementation are working correctly. Ready for your review!

You
October 27, 2025 at 11:20 PM

Thanks for the quick implementation! I've tested the login flow and it works great. Can you please add the password strength validation and remember me checkbox as mentioned in the acceptance criteria?

And that's pretty much it! There are more you can do with Agiflow, but with this simple workflow, you can now easily scale out multiple projects all at once and leverage the full power of your subscription.

Installation

Get the Agiflow CLI installed and ready to connect with your AI agents.

1. Install the Agiflow CLI

The Agiflow CLI manages agent connections, session tracking, and model routing.

# Install globally with npm
npm install -g @agiflowai/agent-cli
# Or use with npx (no installation)
npx @agiflowai/agent-cli

2. Verify Installation

Check that the CLI was installed correctly:

# Check version
agent-cli --version

Connect & Launch Agents

1. Connect to Agiflow (Optional)

For backend-connected mode, authenticate your CLI with Agiflow:

# Connect daemon to Agiflow backend
agent-cli connect

💡 Tip: This opens your browser for device code authentication. The daemon receives task assignments from the dashboard and routes them to agents.

2. Launch Claude Agent

Start Claude with session tracking and optional model routing:

# Launch Claude for a task
agent-cli claude --agent-session-id LXX-4 --alias "backend-oauth"
# Launch with model routing (GPT-5)
agent-cli claude --llm-provider chatgpt --llm-model gpt-5 --alias "review"
# Standalone mode (no backend)
agent-cli claude --standalone

3. Dynamic Model Routing (Optional)

Change AI models for active sessions without restart:

# Interactive routing menu
agent-cli router

Select session → Choose routing: Claude, GPT-5 (high/medium/low/minimal reasoning), GPT-5 Codex

Example Workflow

Try this workflow with Work Units:

1

Create a Work Unit for a feature

Group 5-8 related tasks under a Work Unit (e.g., "Shopping Cart Feature")

2

Assign agent members to tasks

Backend dev → API tasks, Frontend dev → UI tasks, Reviewer → code review

3

Run tasks with session continuity

Each task updates the Work Unit's devInfo metadata for context-aware execution

Next Steps

Work Units

Learn how Work Units solve AI session state loss with devInfo metadata.

View Work Units

Multi-Agent

Understand agent members, session management, and model routing.

Explore Multi-Agent

Need Help?

Having trouble getting started? Here are some resources to help you:

Create AccountContact Support