Skip to main content

Getting Started

New to AGI Agentic Phone MCP? Start here:

Quickstart

Connect a client and run your first phone task

Connect your AI client

Claude Code, Claude Desktop, Cursor, ChatGPT, IDE extensions

Architecture

How client, cloud, and phone fit together

Task Lifecycle

States, terminal outcomes, and the monitor loop

Core Concepts

Architecture

The three-piece system and the async execution model

Devices & Sessions

Devices, the default device, sessions, and the execution lane

Task Lifecycle

Four states, nine terminal outcomes, human-in-the-loop handoffs

Use Cases & Examples

Real workflows you can run today:

Messaging

Send messages with confirmation gates

Daily Assistant

Calendar, inbox, and multi-app catch-ups

Cross-App Workflows

Multi-task workflows in one session

App Testing & QA

Real-device app walkthroughs and bug reports

Best Practices

Task Design

Write prompts that phone agents execute reliably

Session Management

Sessions, devices, and workflow boundaries

Error Handling

Handle every terminal outcome deliberately

Performance

Timeouts, queue waits, and task sizing

Security

The trust model and how to stay inside it

Monitoring

Steps, snapshots, and debugging live tasks

Troubleshooting

Troubleshooting Guide

Start here: symptom-based index

Devices & Connectivity

Missing devices, offline devices, empty tool lists

Tasks & Sessions

Stuck, paused, pending, and cancelled tasks

Authentication

OAuth and API key failures

MCP Clients

Connection, transport, and client-specific issues

Quick Reference

Tool Reference

Every tool, parameter, and outcome

Agentic Phone MCP Setup

Server details and client setup

Support

Get help from our team