Skip to content

AutoGPT

Open Source

AutoGPT is a cloud platform and open-source project that pairs a coordinating AI agent named Otto with specialized AI Experts for content, sales, research, and recurring operations work. Users brief the team in plain English, connect their tools, and let Otto route tasks to the right specialist.

314

Overview

AutoGPT is a platform that gives individuals and small teams a coordinated set of AI agents—called Experts—led by a generalist agent named Otto, aimed at content, sales, research, and the recurring work that never quite goes away. The pitch is organizational rather than conversational: instead of one chatbot you prompt repeatedly, you assemble specialists, brief them in plain English, and let Otto decide who handles what.

How the Expert model works

The roster currently lists 33 Experts, each framed around a recognizable job. Maria handles SEO content, producing briefs, drafts, and page fixes. Max works as an SDR, assembling prospect lists and first-draft outreach. Jules turns work you have already published into LinkedIn posts, and Kai acts as an executive assistant, sorting inboxes and prepping meetings. Frankie appears in the product's morning-brief example as another assistant role. The framing matters because it sets expectations: these are scoped roles with defined outputs, not a single model asked to be everything.

Otto as the coordinating layer

Otto sits at the head of the team and functions as the entry point for most work. You can tell it what you want done, ask it to involve a specific Expert, or catch up on what the team has produced. The platform describes Otto as capable of generating workflows from a plain-English description, asking clarifying questions, then building and saving the result. Newly generated workflows run through a dry-run self-repair loop that simulates execution and catches errors before they reach you. Otto also tracks structured task lists mid-job, displays reasoning traces, and compacts long conversations to stay coherent.

Building and scheduling workflows

Beyond chat, AutoGPT keeps a visual workflow builder for creating and inspecting automations step by step. Scheduled workflows and recurring routines let work run on a cadence rather than on demand, and each Expert can be given its own spending cap. The platform supports multimodal inputs such as images and PDFs, voice input and output, and a persistent cloud sandbox for running code and managing files. A public API with an auto-generated OpenAPI schema and generated SDK clients gives developers a path to deploy and retrieve workflow results programmatically.

Open-source roots and self-hosting

The project began as an open-source experiment that drew wide press coverage in 2023 from outlets including WIRED, Forbes, Bloomberg, TechCrunch, Fast Company, and Ars Technica, and its GitHub repository has accumulated more than 187,000 stars. That heritage remains practical: you can self-host AutoGPT for free with a single Docker command, supplying your own infrastructure and model access. Self-hosting is supported on Linux, macOS, and Windows via WSL 2, on both x86 and Arm machines. Features that depend on third-party services—model providers, sandboxes, browser automation, email, and similar—require your own API keys when self-hosted.

Control, cost, and the human in the loop

AutoGPT gives users meaningful levers over what each Expert can do. You set an Expert's instructions, choose which connected accounts it may use, install its workflows, and pause scheduled work at any time. Spending controls apply per Expert, and results and requests are reviewable. The company is direct that AI output still needs human judgment, which is a fair framing for a product built around delegation. The main cost consideration is that workflow runs draw on a credit wallet on every plan, so automation volume, not just the subscription, drives the bill.

Where it fits

AutoGPT suits people who already know which tasks they want off their plate and can describe them clearly. The Expert structure shortens the distance between a job title and a working agent, and Otto keeps multi-step projects from fragmenting across separate chats. Teams that need shared workspaces, admin roles, and centralized billing will have to wait—the Team plan is listed as coming soon—while developers who want full control can bypass the hosted service entirely through self-hosting or the API.

Core features

  • Otto, a Head of AI agent that coordinates Experts, answers questions, and builds or runs workflows from plain-English prompts
  • A roster of 33 AI Experts, including an SEO content manager, an SDR, a LinkedIn writer, and an executive assistant
  • Visual workflow builder for creating and inspecting automations step by step
  • Scheduled workflows and recurring routines with per-Expert spending controls
  • Access to a broad range of leading AI models, plus document and connected-app support
  • Multimodal chat inputs, voice in and out, and a persistent cloud sandbox for code and files
  • Self-hosting via a single Docker command with your own infrastructure and model keys
  • Public API with an OpenAPI schema and generated SDK clients

Use cases

  1. 1Drafting SEO briefs, articles, and page fixes through the Maria Expert
  2. 2Building prospect lists and first outreach drafts with the Max SDR Expert
  3. 3Turning existing work into LinkedIn posts via the Jules Expert
  4. 4Sorting an inbox and prepping meetings with the Kai executive assistant Expert
  5. 5Scheduling recurring research and operations routines that run on a credit budget
  6. 6Self-hosting the open-source platform for teams that need to manage their own infrastructure

Strengths and things to check

Strengths

  • Otto gives multi-step work a single coordinating point instead of scattered chats
  • Specialized Experts map to recognizable job functions, which makes delegation intuitive
  • Per-Expert spending caps and reviewable results keep automated runs accountable
  • The open-source core and Docker self-hosting offer an escape hatch from the hosted service
  • A visual builder and API cover both non-technical and developer workflows

Things to check

  • Workflow runs draw on a credit wallet on every plan, so heavy automation adds variable cost beyond the subscription
  • AI output still requires human judgment, and the company says as much on its own site
  • The Team plan with multi-user workspaces and admin roles is listed as coming soon
  • Self-hosting shifts infrastructure, model access, and third-party service costs onto you

Pricing

Pro costs $50/month, or $510/year (equivalent to $42.50/month) with annual billing. Eligible new accounts get a 7-day free trial; a card is required and there is no subscription charge today. Max costs $320/month and includes 8.5x Pro usage, 5x file storage, early access to features, and priority support. A Team plan is listed as coming soon. Workflow runs draw on a credit wallet on every plan. Prices are subject to change.

Basic information

Website
agpt.co
Added
Oct 18, 2026
Platforms
Web
Chinese & English
Not marked
API
Marked as supported
Self-hosting
Marked as supported
Best for
Individuals and small teams who want to delegate recurring content, sales, research, and operations work to a coordinated set of AI agents rather than a single chat assistant.
Information updated
Nov 2, 2026
Website status
Checking
Corrections · Privacy · Data requests?
Contact us

Features, pricing and availability may change. Always verify details on the official website. Maintained by steven · Last updated Nov 2, 2026 →

Related tools

All categories