AutoGPT
FreemiumYour team of AI Experts.
Published 27 September 2026 · Last updated 1 October 2026
Scores
Popularity4/5
Went viral in spring 2023 as the first autonomous agent to reach mainstream tech coverage, and it still has one of the largest GitHub followings of any AI project. Newer open-source agents, OpenClaw and Hermes Agent among them, have since passed it in both stars and attention.
Learning Curve2/5
AutoPilot's chat-based creation removes setup friction for simple agents entirely, though the visual builder and more advanced multi-block workflows take real familiarity to use well.
Flexibility4/5
A drag-and-drop builder, 45+ integrations, hundreds of supported models, and a public marketplace of extendable community agents give it broad configurability for general automation tasks.
Performance3/5
Solid for the continuous-workflow-automation use cases it's now built around, though as a general low-code agent platform rather than a codebase-scoped coding tool, its depth on pure software-engineering tasks trails purpose-built coding agents.
Portability3/5
Split licensing (MIT for the classic agent code, source-available Polyform Shield for the core platform) and a hosted-first product design make it less freely self-hostable at full capability than a fully MIT-licensed alternative.
About AutoGPT
AutoGPT began in 2023 as an open-source experiment that popularized the autonomous AI agent: a program that pursues a goal by planning sub-tasks, executing them, checking progress, and adjusting without step-by-step direction. That original command-line agent survives as AutoGPT Classic. The product today is the AutoGPT Platform, a hosted and self-hostable service for AI agents that carry out recurring business work.
The platform is organized as a team. Otto, a general-purpose chat agent, takes a brief in plain language, delegates to specialists called Experts, and reports back. Experts are ready-made agents for roles such as SEO content, sales prospecting, bookkeeping, and customer support, each with its own instructions, skills, and workflows, and you can define your own.
Underneath sits a visual workflow builder: a drag-and-drop canvas where blocks for models, apps, web search, browser automation, and code execution are wired together with branches and loops. Workflows run on demand, on a schedule, or from a webhook trigger, and a community marketplace offers published workflows to run or fork. Integration blocks cover tools such as Gmail, Slack, GitHub, Notion, and HubSpot, and MCP servers extend the list.
The hosted service is a paid subscription that covers chat with Otto and the Experts, with workflow runs billed separately from a prepaid credit wallet. Self-hosting with Docker is free, using your own model API keys and infrastructure. Licensing is split: the platform code is under Polyform Shield, free for personal and internal business use but not for resale as a competing hosted service, while AutoGPT Classic remains MIT.
Key Features
- Otto, a chat agent that takes a brief, delegates to Experts, and builds workflows from a prompt
- Ready-made Experts for marketing, sales, finance, support, operations, and research roles, plus custom ones you define
- Visual drag-and-drop workflow builder with branching, loops, and nested workflows
- Workflows run on demand, on a cron schedule, or from webhook triggers
- Marketplace of community-built workflows to run, save, or fork
- 200+ integration blocks, including Gmail, Slack, GitHub, Notion, HubSpot, and MCP servers
- Models from Anthropic, OpenAI, Google, xAI, Meta, and Mistral, with no provider keys needed on the hosted service
- Per-run cost display and a spending cap for each Expert
Pros
- Ready-made Experts give a working starting point without designing a workflow first
- Any workflow can be opened on the visual canvas for block-by-block control
- The hosted service supplies model access and most integration credentials, so there are no provider API keys to manage
- Free self-hosting with Docker for teams that want to run it on their own infrastructure
Cons
- No free hosted plan: after the 7-day trial the hosted service needs a paid subscription
- Hosted costs come in two parts, a subscription for chat and a credit wallet for workflow runs, which makes spend harder to predict
- The platform license (Polyform Shield) is source-available with limits on competing hosted services, not the MIT license the original project used
- Aimed at business tasks such as content, sales, and operations, so it is a weaker fit for work inside a codebase than Claude Code or Cursor
- Newer open-source agents such as OpenClaw and Hermes Agent have passed it in GitHub stars and community attention
AutoGPT Pricing
Freemium- · Run the platform yourself with your own API keys and infrastructure
- · $42.50 a month billed annually
- · 7-day free trial
- · Otto and Expert chat, the visual workflow builder, and scheduled workflows
- · Workflow runs billed separately from a credit wallet
- · $272 a month billed annually
- · 8.5x the usage of Pro and 5x the file storage
- · Priority support and early feature access
- · Coming soon; contact sales
- · Multi-user workspaces, admin controls, and centralized billing
Tools Related to AutoGPT
Integrates with AutoGPT(3)
Claude is the default model behind Otto, AutoGPT's chat agent, and a built-in provider for the AI blocks in a workflow. The hosted service includes Claude access; self-hosters add their own Anthropic API key.
OpenAI is a built-in model provider in AutoGPT: GPT models can be picked in any AI block of a workflow. The hosted service includes access; self-hosters add their own OpenAI API key.
A self-hosted AutoGPT can run its AI blocks on local models through the built-in Ollama provider: point a block at your Ollama server and no model API key is needed.
Alternatives to AutoGPT(10)
n8n is a general workflow automation tool with hundreds of app integrations and an AI Agent node; AutoGPT starts from the agent, with ready-made Experts and a chat agent that builds workflows from a brief. Both can be self-hosted. Pick n8n to connect many business apps, AutoGPT when the agent does the work.
AutoGPT and Manus are both general-purpose, non-codebase-scoped agent platforms focused on business tasks. AutoGPT is open-source and self-hostable with a visual builder and marketplace; Manus is a closed-source managed cloud product focused on producing finished deliverables from a single high-level goal.
Hermes Agent is a personal agent that learns reusable skills from its own experience and runs self-hosted or on Hermes Cloud; AutoGPT is a business platform of role-based AI Experts and visual workflows. Pick Hermes for one assistant that improves over time, AutoGPT for scheduled business workflows.
Langflow is a free, MIT-licensed canvas for wiring models, prompts, tools, and vector stores into agent and RAG pipelines served as an API; AutoGPT pairs its workflow canvas with ready-made business Experts and a hosted subscription. Pick Langflow to build a pipeline into your own product, AutoGPT for business tasks run for you.
Dify is a platform for building your own workflows, RAG chatbots, and agents and publishing them as apps or APIs; AutoGPT pairs a workflow canvas with ready-made business Experts and a hosted subscription. Pick Dify to build AI apps for others to use, AutoGPT for business tasks run for you.
AutoGPT and Grok Bot both target business-operations automation rather than coding. AutoGPT is an open-source, self-hostable platform with a visual builder and marketplace; Grok Bot is xAI's paid, cloud-only "AI teammates" product bundled into business subscriptions.
Vendor
Significant Gravitas
Website →Tags
Details
- Maintained
- Yes
- Agent type
- Autonomous app
- Primary language
- Python
- Open source
- Yes
- Hosting
- Cloud & Self-hosted
- GitHub stars
- 188k
- Stars updated
- 2026-10-01