Skip to main content

​
Overview

Production-ready autonomous agents that take action for you. The AGI Agent Sessions API enables you to create browser agent sessions that can:

Internet Research

Gather information, compare options, and compile reports

Form Automation

Fill out forms, complete registrations, and handle workflows

E-commerce

Make purchases, track prices, and monitor inventory

Web Automation

Navigate complex sites and complete multi-step tasks
Trusted by leading companies across agentic commerce, GTM research, sales, marketing, and productivity automation.

​
Base URL

All API requests should be made to: 
https://api.agi.tech/v1

​
Authentication

Secure by default. All API endpoints require authentication using a Bearer token. Include your API key in the Authorization header of every request: 
const headers = {
    'Authorization': `Bearer ${apiKey}`,
    'Content-Type': 'application/json'
};
Security Note: If authentication fails, endpoints will return 401 (Unauthorized) or 403 (Forbidden) status codes. Keep your API keys secure and never commit them to version control.

​
Getting Started

Five simple steps to deploy your first agent:
1

Create a session

Start by creating a new agent session with your preferred model
2

Send a message

Give the agent a task using natural language instructions
3

Monitor progress

Poll for status updates or stream events in real-time via SSE
4

Control execution

Pause, resume, or cancel tasks as needed
5

Clean up

Delete the session when you’re done

Quickstart Guide

Get your first agent running in under 5 minutes

​
Key Concepts

​
Sessions

A session represents a single browser environment with an agent. Each session has its own isolated browser state and can execute one task at a time.

Session Isolation

Each session runs in a completely isolated environment for security and reliability

​
Agent Models

Choose the right agent model for your use case:

agi-0

Default model with full capabilities and maximum intelligence

agi-0-fast

Optimized for speed with faster response times

​
Message Types

The agent communicates through structured messages:
Agent’s reasoning process or actions being taken
Questions requiring user input or clarification
Task completion with final results
Error notifications and failure states
Additional context or debugging information

​
Session Status vs Execution Status

Two complementary status systems help you track agent state:
  • Session Status: Lifecycle state (initializing, ready, running, paused, completed, error, terminated)
  • Execution Status: Current task state (running, waiting_for_input, finished, error)

​
Interaction Patterns

Two ways to monitor your agents — choose what works best for your architecture:

​
Polling

Simple and reliable. Use the /messages endpoint with the after_id parameter to poll for new messages: 
const messages = await session.getMessages({ afterId: 0 });
Track the highest message id you’ve seen and use it as after_id in subsequent requests to get only new messages.

​
Server-Sent Events (SSE)

Real-time updates. Connect to the /events endpoint for live streaming: 
for await (const event of session.streamEvents()) {
    console.log(event);
}

SSE Benefits

Lower latency, reduced polling overhead, and true real-time monitoring

​
Security & Risk Prevention

Enterprise-grade security built into every request:

Rate Limiting

Automatic throttling prevents abuse and ensures fair usage

Malicious Intent Screening

Real-time screening before command execution

Secure Payments

Support for 1-time-use payment tokens

Session Isolation

Complete isolation between sessions

​
Support

We’re here to help. For issues, questions, or feature requests:

Company Website

Visit theagi.company
Ready to dive deeper? Explore the detailed endpoint documentation below.